Before DENNIS, ELROD, and DUNCAN, Circuit Judges. JENNIFER WALKER ELROD and STUART KYLE DUNCAN, Circuit Judges: On April 7, 2020, we issued a writ of mandamus vacating the district court’s temporary restraining order (“TRO”)[1] that exempted abortions from GA-09, an emergency measure temporarily postponing non-essential medical procedures during the COVID-19 pandemic. In re Abbott, — F.3d —, 2020 WL 1685929 (5th Cir. Apr. 7, 2020) (Abbott II). Two days later, on April 9, the district court entered a second TRO, exempting various categories of abortion from GA-09. See Planned Parenthood Ctr. for Choice v. Abbott, No. A-20-CV- 323, 2020 WL 1815587 (W.D. Tex. Apr. 9, 2020) (Abbott III). A flurry of litigation ensued, during which state officials again sought mandamus and we administratively stayed parts of the April 9 TRO.[2] Over this period—from April 7 to 20—Texas COVID-19 cases, hospitalizations, and deaths more than doubled.[3] We now consider the mandamus petition directed to the April 9 TRO. We are persuaded by Petitioners’ arguments that the district court, in the April 9 TRO, disregarded our mandate in Abbott II. The court again “fail[ed] to apply . . . the framework governing emergency exercises of state authority during a public health crisis, established over 100 years ago in Jacobson v. Commonwealth of Massachusetts, 197 U.S. 11 (1905).” Abbott II, 2020 WL 1685929, at *5. Moreover, the court again second-guessed the basic mitigation strategy underlying GA-09 (that is, the concept of “flattening the curve”), and also acted without knowing critical facts such as whether, during this pandemic, abortion providers do (or should) wear masks or other protective equipment when meeting with patients. Those errors led the district court to enter an overbroad TRO that exceeds its jurisdiction, reaches patently erroneous results, and usurps the state’s authority to craft emergency public health measures “during the escalating COVID-19 pandemic.” Id. at *1. Once again, the dissenting opinion accuses the majority of treating abortion differently and once again it is wrong. At issue is whether abortion can be treated the same as other procedures under GA-09. It is the district court that treated abortion differently, issuing back-to-back TROs that did not follow the law. We therefore grant the writ in part and direct the district court to vacate these parts of the April 9 TRO: That part restraining enforcement of GA-09 as a “categorical ban on all abortions provided by Plaintiffs.” That part restraining the Governor of Texas and the Attorney General. That part restraining enforcement of GA-09 as to medication abortions. That part restraining enforcement of GA-09 as to patients who would reach 18 weeks LMP[4] on the expiration date of GA-09 and who would be “unlikely” to be able to obtain abortion services in Texas. That part restraining enforcement of GA-09 after 11:59 p.m. on April 21, 2020. We do not grant the writ, and therefore do not order vacatur, of that part of the TRO restraining GA-09 as to patients “who, based on the treating physician’s medical judgment, would be past the legal limit for an abortion in Texas—22 weeks LMP—on April 22, 2020.” I. We summarize the pertinent background, which we have chronicled in greater detail elsewhere. See Abbott II, 2020 WL 1685929, at *1-4; Abbott IV, 2020 WL 1844644, at *1-2. GA-09 is an emergency public health measure, issued by the Governor of Texas on March 22, 2020, that postpones non- essential surgeries and procedures until April 22 to combat the COVID-19 pandemic. It applies to all licensed healthcare providers in Texas, covers a broad range of procedures, does not mention abortion, and contains life-and-health exceptions committed to a physician’s judgment. Specifically, GA-09 requires healthcare professionals and facilities to: postpone all surgeries and procedures that are not immediately medically necessary to correct a serious medical condition of, or to preserve the life of, a patient who without immediate performance of the surgery or procedure would be at risk for serious adverse medical consequences or death, as determined by the patient’s physician.[5] The order does not apply to procedures that, if performed under accepted standards, “would not deplete the hospital capacity or the personal protective equipment ["PPE"] needed to cope with the COVID-19 disaster.”[6] GA-09 is enforceable by criminal and administrative penalties and expires at 11:59 p.m. on April 21, 2020.[7] See Abbott II, 2020 WL 1685929, at *2-4 & nn.10-12. When ordering vacatur of the first TRO, we explained that Respondents’ challenge to GA-09 must satisfy the standards in Jacobson v. Massachusetts, 197 U.S. 11 (1905). Specifically, we held: [W]hen faced with a society-threatening epidemic, a state may implement emergency measures that curtail constitutional rights so long as the measures have at least some “real or substantial relation” to the public health crisis and are not “beyond all question, a plain, palpable invasion of rights secured by the fundamental law.” Jacobson, 197 U.S. at 31. Courts may ask whether the state’s emergency measures lack basic exceptions for “extreme cases,” and whether the measures are pretextual—that is, arbitrary or oppressive. Id. at 38. At the same time, however, courts may not second-guess the wisdom or efficacy of the measures. Id. at 28, 30. Abbott II, 2020 WL 1685929, at *7 (cleaned up). We also articulated how the Jacobson framework works with the Casey undue-burden analysis. Id. at *11 (discussing Planned Parenthood of Se. Pa. v. Casey, 505 U.S. 833 (1992)). A court should “ask[ ] whether GA-09 imposes burdens on abortion that ‘beyond question’ exceed its benefits in combating the epidemic Texas now faces.” Id. (quoting Jacobson, 197 U.S. at 31). We emphasized that this analysis would “require[ ] careful parsing of the evidence” and that “[t]hese are issues that the parties may pursue at the preliminary injunction stage, where Respondents will bear the burden to prove, by a clear showing, that they are entitled to relief . . . in any particular circumstance.” Id. at *11-12 (cleaned up). The day following our mandamus, April 8, 2020, the district court: (1) vacated its March 30 TRO; (2) cancelled the telephonic preliminary injunction hearing previously scheduled for April 13; and (3) ordered the parties to file a joint status report by April 15 outlining a schedule for a new preliminary injunction hearing on a yet-unannounced date. That same day, Respondents filed a new TRO application supported by one new declaration. The next day, April 9, the district court convened a brief telephone conference with the parties, during which the court declined to allow Petitioners either to file a responsive pleading or submit evidence opposing the application. In doing so, the court remarked to Petitioners, “[I]f I were to make a ruling that was unsatisfactory to the State defendants before then, then you may head back to the Circuit with it.” Transcript of 4/9/20 Tele. Conf. at 14:39. Later that day, the court issued a new TRO. Abbott III, 2020 WL 1815587. Adopting Respondents’ proposed findings of fact and conclusions of law, compare id. at *1-7, with Proposed TRO, App. 445-57, this TRO restrains Petitioners from enforcing GA-09 as follows: (1) “as a categorical ban on all abortions provided by Plaintiffs”; (2) as to “medication abortions”; (3) as to “procedural[8] abortion[s] [provided] to any patient who, based on the treating physician’s medical judgment, would be more than 18 weeks LMP on April 22, 2020, and likely unable to reach an ambulatory surgical center in Texas or to obtain abortion care”; and (4) as to “procedural abortion[s] [provided] to any patient who, based on the treating physician’s medical judgment, would be past the legal limit for an abortion in Texas—22 weeks LMP—on April 22, 2020.” Abbott III, 2020 WL 1815587, at *7. On April 10, Petitioners again requested mandamus from our court, this time seeking vacatur of the April 9 TRO. On April 10, we administratively stayed the TRO except as to women who would reach 22 weeks LMP on April 22. Abbott IV, 2020 WL 1844644. On April 13, we denied an emergency stay, and lifted the administrative stay, as to that part of the TRO applying to medication abortions. Abbott V, 2020 WL 1866010.[9] On April 14, the district court set a telephonic preliminary injunction hearing for April 29. Doc. 82. The court also extended the April 9 TRO—”in its entirety under the same terms and conditions except as modified by [our orders]“—until May 1, 2020, at 5:00 p.m. Id. The court stated there was “good cause” for extending the TRO “so that the court and parties have adequate time to prepare for [the April 29] hearing.” Id. On April 15, the Governor issued executive order GA-15, which becomes effective when GA-09 expires and continues until 11:59 p.m. on May 8, 2020. GA-15[10] is similar to GA-09, but has some textual differences as well as an additional exception for certain facilities.[11] II. Federal courts “may issue all writs necessary or appropriate in aid of their respective jurisdictions and agreeable to the usages and principles of law.” 28 U.S.C. § 1651(a). That includes the writ of mandamus sought by Petitioners. See Cheney v. U.S. Dist. Court for D.C., 542 U.S. 367, 380 (2004); In re Gee, 941 F.3d 153, 157 (5th Cir. 2019). Mandamus is proper only in “exceptional circumstances amounting to a judicial usurpation of power or a clear abuse of discretion.” In re Volkswagen of Am., Inc., 545 F.3d 304, 309 (5th Cir. 2008) (en banc) (quoting Cheney, 542 U.S. at 380). An “abuse of discretion” becomes a “clear abuse of discretion” when it “produce[s] a patently erroneous result.” Id. at 310. The writ has issued “where it was the only means of forestalling intrusion by the federal judiciary on a delicate area of federal-state relations [and] where it was necessary to confine a lower court to the terms of an appellate tribunal’s mandate.” Will v. United States, 389 U.S. 90, 95-96 (1967) (citing Maryland v. Soper, 270 U.S. 9 (1926) (federal-state relations) and United States v. U.S. Dist. Ct., 334 U.S. 258 (1948) (effectuating appellate mandate)). Before prescribing this strong medicine, “[w]e ask (1) whether the petitioner has demonstrated that it has no other adequate means to attain the relief it desires; (2) whether the petitioner’s right to issuance of the writ is clear and indisputable; and (3) whether we, in the exercise of our discretion, are satisfied that the writ is appropriate under the circumstances.” In re Itron, Inc., 883 F.3d 553, 567 (5th Cir. 2018) (quoting Cheney, 542 U.S. at 380-81) (cleaned up). “These hurdles, however demanding, are not insuperable. They simply reserve the writ for really extraordinary causes.” Gee, 941 F.3d at 158 (cleaned up). In such cases, mandamus provides a “useful ‘safety valve[ ]‘ for promptly correcting serious errors.” Mohawk Indus., Inc. v. Carpenter, 558 U.S. 100, 111 (2009) (quoting Digital Equipment Corp. v. Desktop Direct, Inc., 511 U.S. 863, 883 (1994)). As in Abbott II, we address each prong in turn, beginning with the second. Abbott II, 2020 1685929, at *5. III. A. Failure to Narrowly Tailor April 9 TRO We first address two threshold errors in the April 9 TRO that demonstrate Petitioners’ right to the writ. Because “the scope of injunctive relief is dictated by the extent of the violation established, [t]he district court must narrowly tailor an injunction to remedy the specific action which gives rise to the order.” John Doe #1 v. Veneman, 380 F.3d 807, 818 (5th Cir. 2004) (cleaned up). The April 9 TRO fails this narrow tailoring requirement in two obvious ways. First, the TRO enjoins enforcement of GA-09 “as a categorical ban on all abortions provided by Plaintiffs.” Abbott III, 2020 WL 1815587, at *7. But GA- 09 is obviously not a “categorical ban on all abortions.” Because it expires on April 22, it is not a ban, but a generally applicable postponement of PPE- consuming surgeries and procedures. And as we have explained already, GA- 09 facially exempts surgeries and procedures immediately necessary to “correct a serious medical condition of, or to preserve the life of, a patient who without immediate performance of the surgery or procedure would be at risk for serious adverse medical consequences or death, as determined by the patient’s physician.” Abbott II, 2020 WL 1685929, at *3. The district court reached its overbroad construction of GA-09 by referring to the Attorney General’s “interpretation” in a “press release,” which the court maintained “has been adopted by the State Defendants.” Abbott III, 2020 WL 1815587, at *2. But Abbott II already found this chain of reasoning flawed. We found “no reason to believe [the] press release has the force of law,” and, in any event, the press release itself recognized GA-09 exempts abortions “medically necessary to preserve the life or health of the mother.” Abbott II, 2020 WL 1685929, at *13 n.25. The district court also cited no evidence suggesting that the “State Defendants” have adopted its overreading of GA-09. Second, as now extended to May 1, the April 9 TRO is not “narrowly tailor[ed]” to remedy any harm caused by GA-09 because it extends beyond the expiration of GA-09. See John Doe #1, 380 F.3d at 818. By its terms, GA-09 lasts “until 11:59 p.m. on April 21, 2020.”[12] After that point, there will be no “actual case or controversy” between the parties, John Doe #1, 380 F.3d at 814 (citation omitted), and no enforcement of GA-09 for a court to restrain. The fact that the Governor has since announced that a new order—GA-15—will take effect on April 22 does nothing to change this conclusion, as the extended TRO at issue here applies only to GA-09. By purporting to restrain Petitioners past the expiration date of GA-09, the district court exceeded its jurisdiction. See, e.g., Preiser v. Newkirk, 422 U.S. 395, 401 (1975) (a federal court’s judgment must award “specific relief through a decree of a conclusive character, as distinguished from an opinion advising what the law would be upon a hypothetical state of facts”) (quoting North Carolina v. Rice, 404 U.S. 244, 246 (1971)). Likewise, “since the scope of injunctive relief is dictated by the extent of the violation established,” the relief was overbroad because no violation can occur after 11:59 p.m. on April 21. See Califano v. Yamasaki, 442 U.S. 682, 702 (1979). We therefore conclude that Petitioners have demonstrated entitlement to the writ. B. Failure to Dismiss Governor and Attorney General Under Eleventh Amendment Petitioners also argue they are entitled to mandamus relief because the district court violated the Eleventh Amendment by purporting to enjoin the Governor and Attorney General. We agree. In Abbott II, we instructed the district court to “consider whether the Eleventh Amendment requires dismissal of the Governor or Attorney General because they lack any ‘connection’ to enforcing GA-09 under Ex parte Young, 209 U.S. 123 (1908).” Abbott II, 2020 WL 1685929, at *5 n.17 (citing City of Austin v. Paxton, 943 F.3d 993, 999 (5th Cir. 2019); Morris v. Livingston, 739 F.3d 740, 745-46 (5th Cir. 2014)). The district court’s cursory analysis of this question in its April 9 TRO was wrong. Ex parte Young allows suits for injunctive or declaratory relief against state officials, provided they have sufficient “connection” to enforcing an allegedly unconstitutional law. City of Austin, 943 F.3d at 997 (citing Raj v. La. State Univ., 714 F.3d 322, 328 (5th Cir. 2013)). Otherwise, the suit is effectively against the state itself and thus barred by the Eleventh Amendment and sovereign immunity. See Va. Office for Prot. & Advocacy v. Stewart, 563 U.S. 247, 253 (2011); Will v. Mich. Dep’t of State Police, 491 U.S. 58, 71 (1989); Edelman v. Jordan, 415 U.S. 651, 663-69 (1974). If the official sued is not “statutorily tasked with enforcing the challenged law,” then the requisite connection is absent and “our Young analysis ends.” City of Austin, 943 F.3d at 998 (citing Morris, 739 F.3d at 746). As to the Governor, the district court reasoned he has “some connection” to GA-09 because of his “statutory authority [under] Texas Government Code § 418.012.” Abbott III, 2020 WL 1815587, at *6 (cleaned up). But the cited section empowers the Governor to “issue,” “amend,” or “rescind” executive orders, not to “enforce” them. Tex. Gov’t Code § 418.012. The power to promulgate law is not the power to enforce it. Cf. Martin v. Occupational Safety & Health Review Comm’n, 499 U.S. 144, 152 (1991) (distinguishing between the Secretary of Labor’s “powers to promulgate and to enforce national health and safety standards”). The April 9 TRO addresses only “enforcing” GA-09 against plaintiffs who provide certain abortion procedures. Abbott III, 2020 WL 1815587, at *7. And we have already explained that violating GA-09 may result in “administrative or criminal penalties,” Abbott II, 2020 WL 1685929, at *3 n.12, enforced by health and law enforcement officials and not the Governor. Consequently, we hold the Governor lacks the required enforcement connection to GA-09 and may not be sued for injunctive relief under the Eleventh Amendment. See Morris, 739 F.3d at 746 (when challenged law “does not specially task [Texas] Governor . . . with its enforcement, or suggest that he will play any role at all in its enforcement,” Governor “is not a proper defendant”). As to the Attorney General, the district court reasoned that he has “authority” to prosecute violations of GA-09 “at the request of local prosecutors,” and that he has also “publicly threatened enforcement” against abortion providers. Abbott III, 2020 WL 1815587, at *6. Neither rationale establishes the Attorney General’s “connection” to enforcing GA-09 for Ex parte Young purposes. Nothing in GA-09 tasks the Attorney General with enforcing it. Speculation that he might be asked by a local prosecutor to “assist” in enforcing GA-09, see Tex. Gov’t Code § 402.028, is inadequate to support an Ex parte Young action against the Attorney General. See City of Austin, 943 F.3d at 1000 (evidence that Attorney General “might . . . bring a proceeding to enforce” the law insufficient under Ex parte Young). Nor does a “press release” by the Attorney General, Abbott III, 2020 WL 1815587, at *2, show authority to enforce GA-09 for Ex parte Young purposes. Here, the Attorney General did not even threaten to enforce GA-09 in the disputed press release. The release warns that “[t]hose who violate the governor’s order will be met with the full force of the law.” App. 31. The Attorney General threatened that GA-09 would be enforced, not that he would enforce it. Moreover, our cases do not support the proposition that an official’s public statement alone establishes authority to enforce a law, or the likelihood of his doing so, for Young purposes. Cf., e.g., City of Austin, 943 F.3d at 1001 (applying Ex parte Young exception because Attorney General sent “threatening letters” to enforce DTPA and was authorized to enforce that law) (discussing NiGen Biotech, LLC v. Paxton, 804 F.3d 389, 392-95 (5th Cir. 2015)). Consequently, we hold the Attorney General also lacks the required enforcement connection to GA-09 and may not be sued for injunctive relief under the Eleventh Amendment. See City of Austin, 943 F.3d at 1002. Mandamus is appropriate to “control jurisdictional excesses,” Gee, 941 F.3d at 158 (citation omitted), such as allowing suits against state officials in violation of the Eleventh Amendment and sovereign immunity. See, e.g., Block v. Tex. Bd. of Law Examiners, 952 F.3d 613, 617 (5th Cir. 2020) (“Under the Eleventh Amendment, federal courts lack jurisdiction over suits against nonconsenting states.”); Sissom v. Univ. of Tex. High Sch., 927 F.3d 343, 347 (5th Cir. 2019) (“[B]ecause the Eleventh Amendment textually divests federal courts of jurisdiction over states, it is indispensable to assessing this court’s jurisdiction.”). Petitioners have demonstrated a clear and indisputable right to the writ on this ground. C. Failure to Follow Abbott II Mandate Petitioners are also entitled to mandamus because the district court, in entering the April 9 TRO, failed to follow our mandate in Abbott II. Most obviously, we instructed the district court to analyze GA-09 under “the framework governing emergency exercises of state authority during a public health crisis, established . . . in Jacobson.” Abbott II, 2020 WL 1685929, at *5. We articulated the Jacobson framework, id. at *6-7, and emphasized that adhering to its narrow compass of judicial review is necessary to prevent courts from “second-guess[ing] the wisdom or efficacy” of emergency public health measures. Id. at *7. Yet the district court did not apply Jacobson: indeed, the court did not even state what Jacobsons framework is, but instead merely cited Jacobson in passing in its conclusion. See Abbott III, 2020 WL 1815587, at *6 (stating only that applying GA-09 to certain abortion categories “violates the standards set forth in both [Casey] and [Jacobson]).” That flatly contradicted our Abbott II mandate, which left no doubt that “[o]ur overriding consideration” was that any further proceedings “adhere to the controlling standards” in Jacobson “for adjudging the validity of emergency measures like [GA-09].” Abbott II, 2020 WL 1685929, at *2. The April 9 TRO violated the “mandate rule,” a particular manifestation of the law-of-the-case doctrine barring reexamination of issues already decided by an appellate court. See United States v. Smith, 814 F.3d 268, 273 (5th Cir. 2016). Under the mandate rule, a district court “must implement both the letter and the spirit of the appellate court’s mandate and may not disregard the explicit directives of that court.” United States v. Lee, 358 F.3d 315, 321 (5th Cir. 2004) (quoting United States v. Matthews, 312 F.3d 652, 657 (5th Cir. 2002)). Thus, this court has held that a district court violated the mandate rule when, after an appeal, a district court modified a consent decree “without holding a hearing and demanding a more developed factual record.” League of United Latin Am. Citizens, Dist. 19 v. City of Boerne, 675 F.3d 433, 437-38 (5th Cir. 2012) (LULAC II). Where a district court fails to fully implement the mandate, a party may seek a writ of mandamus to enforce compliance. See Kapche v. City of San Antonio, 304 F.3d 493, 500 (5th Cir. 2002). Our Abbott II opinion plainly expected, as a foundational premise for applying Jacobson, that the district court would allow the parties to adduce additional evidence about the effects of GA-09 in specific circumstances. Our opinion made this impossible to miss. For example, we said that “[t]he district court has scheduled a telephonic preliminary injunction hearing for April 13, 2020, when all parties will presumably have the chance to present evidence on the validity of applying GA-09 in specific circumstances.” Abbott II, 2020 WL 1685929, at *2 (emphases added). Following that adversarial hearing, we explained, “[t]he district court can then make targeted findings . . . about the effects of GA-09 on abortion access.” Id. (emphasis added). We said the same thing a few pages later: despite finding no evidence in the record that GA-09 violated Casey, we stated that “Respondents will have the opportunity to show at the upcoming preliminary injunction hearing that certain applications of GA-09 may” violate Casey “if they prove that, ‘beyond question,’ GA-09′s burdens outweigh its benefits in those situations.” Id. at *9 (first and third emphases added). Similarly, after canvassing the record, we declined to decide whether a more narrowly tailored injunction would satisfy Jacobson because “parties may pursue [those issues] at the preliminary injunction stage,” then scheduled for April 13. Id. at *12 (emphasis added). And again: in assessing whether Respondents had any evidence showing GA-09 pretextually targeted abortion, we found “no evidence . . . [on] the record before us” of pretext, but stated that “Respondents will have the opportunity . . . to present additional evidence in conjunction with the district court’s preliminary injunction hearing.” Id. at *13 (emphasis added). To be sure, the district court could have rescheduled the preliminary injunction hearing (as it now has done, to April 29) or afforded the parties some other way of presenting new evidence on the burdens and benefits of GA-09 in specific circumstances. But our opinion left no doubt that an additional evidentiary showing was necessary to properly apply Jacobson in particular circumstances. Among other gaps in the record, for example, was evidence showing what PPE is being used in medication and surgical abortions during the current pandemic, or evidence showing the standard of care for those procedures during the pandemic. See infra Part III.D.1.a. Without any means of answering critical questions like those, the district court lacked any basis for finding, as Jacobson requires, that GA-09 lacks a “real or substantial relation” to the health crisis, or that “beyond all question” it “plain[ly]” violates Casey. Abbott II, 2020 WL 1685929, at *6 (quoting Jacobson, 197 U.S. at 31). It is no answer to say that a TRO may be based on a one-sided evidentiary record. See Fed. R. Civ. P. 65(b)(1) (allowing issuance of TRO without notice); Dilworth v. Riner, 343 F.2d 226, 229 (5th Cir. 1965) (observing that TROs are “generally issued ex parte or after a hearing of a summary character”). Our plain instructions in Abbott II were that properly applying Jacobson to GA-09 required “additional evidence” targeted to specific circumstances. Abbott II, 2020 WL 1685929, at *13. It is also no answer to say that our decision did not tell the district court not to cancel the preliminary injunction hearing and enter a different TRO. The mandate rule requires the district court to “implement both the letter and the spirit of the appellate court’s mandate.” Lee, 358 F.3d at 321. Our decision mentioned the then- upcoming preliminary injunction hearing seven times as a forum for adducing evidence from both sides about specific applications of GA-09. See Abbott II, 2020 WL 1685929, at *2, *4 n.16, *8 n.19, *9, *11 n.24, *12, *13. The district court flouted both the letter and the spirit of our mandate by cancelling that adversarial hearing, convening a snap-TRO “hearing” at which one side was barred from offering evidence or argument, and then immediately issuing a new TRO based on evidence we had already ruled insufficient to show a violation of Jacobson and Casey. See LULACII, 675 F.3d at 438. The LULAC litigation provides helpful guidance. In LULAC I, this court vacated the modification of a consent decree because “the paucity of the record in [that] case provided an insufficient basis for the district court to determine that modification was warranted.” League of United Latin Am. Citizens, Dist. 19 v. City of Boerne, 659 F.3d 421, 438 (5th Cir. 2011) (LULAC I). This court instructed that on remand, “the district court should permit supplemental filings and conduct proceedings, as necessary, to develop a sufficient record.” Id. at 439-40. Yet on remand, the district court entered a new “temporary” modification of the consent decree, without “permit[ting] the parties to conduct discovery, or hold an evidentiary hearing to receive competing expert and lay testimony, or even offer [one party] a substantial opportunity to rebut the evidence that [the other parties] presented.” LULAC II, 675 F.3d at 438. The LULAC II panel vacated that new “temporary” order, holding that “[b]y approving a modification of the Consent Decree without holding a hearing and demanding a more developed factual record, the district court failed to follow the ‘letter and spirt’ of the LULAC I mandate.” Id. at 438. So too here. After explaining that the factual record was insufficient to support the TRO in Abbott I, we instructed that after the “preliminary injunction hearing scheduled for April 13, 2020″ at which the parties could “present additional evidence,” Abbott II, 2020 WL 1685929, at *13, the district court could find that GA-09 constituted an undue burden if “beyond question” the law’s burdens exceeded its benefits. Id. at *11. “The district court was required to do this analysis” the first time, we explained, and “that analysis would have required careful parsing of the evidence.” Id. Yet on remand the district court entered a second TRO “without holding a hearing and demanding a more developed factual record.” See LULAC II, 675 at 438. In doing so, “the district court failed to follow the ‘letter and spirit’ of the” Abbott II mandate. See id.[13] To be sure, Respondents suggest that the April 9 TRO is based on a “more robust” record than the one on which the district court based its March 30 TRO. But on critical points, which we analyze in more detail below, the April 9 TRO relied on the same ten declarations already before the district court when it issued the March 30 TRO.[14] Furthermore, after the March 30 TRO issued, Respondents filed supplemental declarations in the district court record—and then proceeded to use those declarations to defend against mandamus in our court.[15] In granting mandamus, we reviewed the record—including those supplemental affidavits—and found the record before us failed to support the conclusion that GA-09 violates Jacobson and Casey.[16] The district court hardly answered Abbott IIs call for more evidence by relying on evidence we had already reviewed and found wanting. Moreover, we called for additional evidence from both sides. See Abbott II, 2020 WL 1685929, at *2 (emphasizing “all parties” would be able to “present evidence on the validity of applying GA- 09 in specific circumstances”). Yet the district court barred Petitioners from proffering new evidence or argument with respect to the April 9 TRO. Mandamus is justified to correct the district court’s failure to follow our Abbott II mandate. See, e.g., Will v. United States, 389 U.S. 90, 95-96 (1967) (explaining that “the writ [of mandamus] has been invoked . . . where it was necessary to confine a lower court to the terms of an appellate tribunal’s mandate”). This is all the more vital here because the failure to follow our mandate led the district court to “embarrass the executive arm of the Government” and “intru[de] . . . on a delicate area of federal-state relations.” Cheney, 542 U.S. at 381 (cleaned up).[17] Here too, Petitioners have demonstrated their clear and indisputable right to the writ. D. Patently Erroneous Results and Usurpation of the State’s Authority to Craft Emergency Health Measures Mandamus relief is also justified because the district court’s failure to follow our Abbott II mandate led to patently erroneous results and usurped the state’s authority to craft emergency public health measures. See In re JPMorgan Chase & Co., 916 F.3d 494, 500 (5th Cir. 2019) (cleaned up) (mandamus warranted where there has been a “usurpation of judicial power” or “a clear abuse of discretion that produces patently erroneous results”). We discuss these problems below, both to explain why we grant mandamus as to two of the three categories of abortion procedures restrained by the April 9 TRO, and also to provide guidance at the preliminary injunction stage. 1. The April 9 TRO Patently Erred by Exempting Medication Abortions from GA-09 There is no constitutional right to any particular abortion procedure. Gonzales v. Carhart, 550 U.S. 124, 164-65 (2007). Yet the district court bluntly concluded that GA-09′s temporary postponement of one kind of early-abortion method—medication abortions—is “beyond question” a violation of Casey. See Abbott III, 2020 WL 1815587 at *6 (concluding, “based on the court’s findings of fact, it is beyond question that [GA-09's] burdens outweigh the order’s benefits as applied to . . . medication abortion”). Despite our instructions in Abbott II, the district court failed to compile a record that remotely justifies this conclusion. Indeed, the record before the district court—which we already reviewed in Abbott II and found inconclusive—does not provide the tools even to answer the pertinent factual question. That question is not, as the district court evidently thought, whether medication abortion consumes PPE during normal circumstances, but instead whether it does so under the pandemic conditions Texas faces and GA-09 addresses. As for the legal question, the district court’s analysis fails to address why temporary postponement of one type of early-abortion method is “beyond question” unconstitutional if it leaves open other means of obtaining an abortion. Restraining state officials from implementing an emergency health measure based on such findings is “a clear abuse of discretion that produces patently erroneous results.” Abbott II, 2020 WL 1685929, at *5 (quoting JPMorgan Chase, 916 F.3d at 500 (cleaned up)). a. Failure to consider PPE usage and standard of care during the pandemic As a general matter, we observe that the regulation of medication abortion in Texas differs from some other states. In Texas, “[b]efore the physician gives, sells, dispenses, administers, provides, or prescribes an abortion-inducing drug, the physician must examine the pregnant woman.” Tex. Health & Safety Code § 171.063(c). During that examination, the patient must receive an ultrasound examination. Tex. Health & Safety Code § 171.012(a)(4). The physician cannot provide the patient an abortion until the second visit. Id. And the patient must schedule a follow-up appointment to ensure the abortion is complete. Tex. Health & Safety Code § 171.063(e)-(f); 25 Tex. Admin. Code 139.53(b)(4).[18] The district court found, as a matter of fact, that “[p]roviding medication abortion does not require the use of any PPE.” Abbott III, 2020 WL 1815587, at *3, 1 15. The pertinent question, however, is whether medication abortions require PPE during the COVID-19 pandemic. See GA-09 (stating that “a shortage of hospital capacity or [PPE] would hinder efforts to cope with the COVID-19 disaster”). Respondents submitted no evidence on that question: they neither stated what PPE they were consuming “during the COVID-19 disaster,” nor submitted evidence establishing the standard of care for medication abortions during the pandemic. Scour the twenty declarations Respondents submitted to support their claim. Does any testify that during the current pandemic, abortion providers are not wearing masks? No. Nor would one expect such a statement when everyday life now presents police officers, priests, mail carriers, grocery store cashiers, gas station attendants, and retail clerks wearing them every day.[19] The question, then, is not whether medication abortions consume PPE in normal times, but whether they consume PPE during a public health emergency involving a spreading contagion that places severe strains on medical resources. See Abbott II, 2020 WL 1685929, at *1. The record contains scant material to answer that question—certainly not to a degree to permit the conclusion that merely postponing medication abortions “beyond question” violates the right to abortion. The April 9 TRO did not analyze PPE consumption for medication abortions during the COVID-19 pandemic. The district court, with one minor exception, relied exclusively on declarations that were before it when it issued the March 30 TRO. See Abbott III, 2020 WL 1815587 at *3, 11 10, 13, 15 (relying on prior declarations); but see id. 1 14 (relying on new declaration). In Abbott II, we explained that those declarations were “unclear” as to “how PPE is consumed in medication abortions.” See Abbott II, 2020 WL 1685929 at *11. Those declarations did not, and still do not, speak to the question of PPE usage during the present public health emergency. No. 20-50296 Moreover, there has been no consideration yet how the pandemic has affected the standard of care for abortion. No record evidence supports the contention—which provides the unstated premise of the district court’s findings—that the standard of care for medication abortion during the COVID- 19 is identical to the normal standard. Relatedly, the record does not establish what PPE abortion providers presently use to protect against the spread of the virus. Indeed, some record evidence indicates that reasonable abortion providers would change PPE usage during the pandemic. For instance, the state’s infectious disease expert declared that “[n]ot wearing face masks and other PPE when caring for patients who are not under investigation for COVID 19 . . . exposes health care workers to transmission of infection” from asymptomatic patients. Marier Decl. 1 12, App. 242. The declarations the district court cited (which are exclusively those of Respondents) consider medication abortion only during normal times. Abbott III, 2020 WL 1815587 at *3, 1 15. One physician describes a clinic’s PPE usage during an “average week.” Wallace Decl. 1 12. That says nothing about PPE usage during a pandemic. Cf. Klier Declaration 1 11, App. 110 (“Before the COVID-19 outbreak, Austin Women’s used no PPE for medication abortion.”) (emphasis added). And a declaration recently filed in the district court clarifies that at least one plaintiff began using surgical masks in response to COVID- 19. See Rosenfeld Decl. 1 13 (“Since the COVID-19 outbreak began, Houston Women’s Clinic has . . . provided our staff with surgical masks (not N95 respirators) . . . .”).[20] In sum, the relevant question is not what PPE is consumed during normal times but “during the COVID-19 disaster,” as GA-09 states. Cf. Abbott II, 2020 WL 1685929 at *12 (“[T]he essence of equity is the ability to craft a particular injunction meeting the exigencies of a particular situation.”). The failure even to consider that question—as well as to support its findings with record evidence—was patently erroneous.[21] b. Usurping state authority to craft emergency health measures As we explained before, Jacobson prohibits courts from “usurp[ing] the state’s authority to craft measures responsive to a public health emergency.” See Abbott II, 2020 WL 1685929, at *12. Courts have no authority to ask whether a “particular method [is]—perhaps, or possibly—not the best.” Jacobson, 197 U.S. at 35. Instead, courts may ask only whether the state has acted in an “arbitrary, unreasonable manner.” Id. at 28. During a pandemic emergency, public authorities must make numerous, complex judgment calls. GA-09 addresses one of the most vexing: how to prevent critical strains on medical resources during a surge in contagious disease. Abbott II, 2020 WL 1685929, at *1-2. Respondents have submitted declarations of infectious disease experts who believe GA-09 is profoundly misguided. See, e.g., Bassett Decl. 1 6-8, App. 311; Sharfstein Decl. 1 9-12, App. 280-81. Texas authorities believe, to the contrary, that GA-09 is critical to protect the state’s citizens and has supported that view with its own medical experts. See, e.g., Marier Decl. 1 12, App. 242. The Supreme Court, and this court, have already explained how to resolve such an impasse: “[I]f the choice is between two reasonable responses to a public crisis, the judgment must be left to the governing state authorities.” Abbott II, 2020 WL 1685929, at *12 (citing Jacobson, 197 U.S. at 30); cf. Dep’t of Commerce v. New York, 139 S. Ct. 2551, 2570 (2019) (explaining, in context of different legal standard, that “the choice between reasonable policy alternatives in the face of uncertainty was the Secretary’s to make”). The district court’s findings in support of the April 9 TRO failed to heed this basic constraint on judicial power. In the April 9 TRO, as in the one before, the district court’s weighing of the public interest substituted its own opinion for the judgment of the governing authorities. What we said before applies here: [T]he district court did little more than assert its own view of the effectiveness of GA-09. The district court did not provide any explanation of its conclusion that the public health benefits from an emergency measure like GA-09 are “outweighed” by any temporary loss of constitutional rights. Abbott II, 2020 WL 1685929 at *12 (discussing Abbott I, 2020 WL 1502102 at *3). In the April 9 TRO, the district court concluded in cursory fashion that Plaintiffs and their patients would “suffer irreparable harm” absent a TRO, that the “balance of equities favors Plaintiffs” and that a TRO “serves the public interest.” Abbott III, 2020 WL 1815587 at *6. The court added “that entry of a [TRO] to restore abortion access would serve the State’s interest in public health.” Id. We find the district court’s approach as flawed this time as the last. To begin with, the district court ignored the entire point of a mitigation measure like GA-09. The concept of “flattening the curve” has become all-too- familiar during the pandemic: as applied to GA-09, it means that delaying procedures now may prevent short-term exhaustion of critical medical resources. This is one stated goal of GA-09: it does not prohibit non-essential procedures, it delays them. As its findings show, however, the district court preferred to second-guess this strategy. For instance, the district court found that delaying abortion access “will not conserve PPE or hospital resources” because women will remain pregnant and thus consume more PPE in the long run. See Abbott III, 2020 WL 1815587 at *4, 11 20-23. But that is not a policy choice federal judges are permitted to make during a public health crisis, if ever.[22] Public authorities are entitled to make a different calculation to protect citizens: even if GA-09 may increase consumption of medical resources in the long run, decreasing consumption now will help weather the immediate surge of COVID-19 cases.[23] Instead of re-weighing the state’s cost-benefit calculus, a federal court “must assume that, when [GA-09] was [issued], the [Governor of Texas] was not unaware of these opposing theories, and was compelled, of necessity, to choose between them.” Jacobson, 197 U.S. at 30. The district court patently erred by doing the opposite. See Jacobson, 197 U.S. at 31; Abbott II, 2020 WL 1685929, at *7. Similarly, the district court found that GA-09 did not promote the public health, in part, because some women might travel to other states to obtain abortions. See Abbott III, 2020 WL 1815587 at *5, 1 25. But the evidence shows, as does common sense, that an emergency measure like GA-09 weighs heavily on people suffering all kinds of health issues. One physician declares she has postponed or canceled surgeries for “patients with possible uterine cancer and cervical cancer diagnoses who are in need of surgeries, as well as patients with heavy bleeding who need surgery but where we can temporarily control the bleeding with medication.” Thompson Decl. 1 4, App. 235. It is possible that those patients too may travel to other states to obtain desired procedures. Moreover, evidence that some women travel to other states to receive an abortion does not demonstrate that GA-09 increases the risk of COVID-19 transmission. Such a claim would require comparing the amount of travel that GA-09 has increased with the amount of travel it has reduced. That calculation is uncertain: One respondent provider declares that some women “come from over a hundred miles to receive care at our clinic.” Dewitt-Dick Decl. 1 22, App. 87. Another testifies that patients at her clinic “hail from all over Texas.” Ferrigno Decl. 1 30, App. 95. A court must assume that the public health experts at the Texas Department of State Health Services—not to mention the CDC, the U.S. Surgeon General, and the Centers for Medicare and Medicaid Services— weighed these difficult trade-offs between medical care and public health. Jacobson, 197 U.S. at 30. Federal judges get no vote on the matter. As the Supreme Court instructed: “[N]o court . . . is justified in disregarding the action of the [Governor] simply because in its opinion that particular method was— perhaps, or possibly—not the best.” Jacobson, 197 U.S. at 35 (cleaned up). The district court’s disregard of that command usurped the power of the state in a public health emergency. c. Failure to carefully parse record evidence The April 9 TRO also failed to “careful[ly] pars[e] the evidence,” as instructed by our previous mandate. See Abbott II, 2020 WL 1685929 at *11. For instance, the district court did not discuss, or even cite, a single declaration of submitted by Petitioners.[24] It did not explain why, to take a conspicuous example, it disregarded the declaration of the state’s infectious disease expert. Nor did the district court mention the undisputed evidence that, “[i]f even one person providing care is carrying COVID-19 but not yet symptomatic, the results could be devastating if that person is not equipped with proper PPE.” Abraham Decl. 1 4, App. 225. The district court did not explain whether it disagreed with this statement or thought it was inapplicable to abortion providers. Nor did the district court mention record evidence indicating that N95 masks are now required for surgical abortions to be performed safely. See Harstad Decl. 1 4.[25] We say this, not to make findings ourselves, but to show why the delicate inquiry in this case requires “careful parsing of the evidence.” Abbott II, 2020 WL 1685929 at *11. A scalpel must be employed, not a rubber stamp. Moreover, the district court’s wholesale adoption of Respondents’ proposed findings resulted in findings that are not supported by the record. One example may suffice. The district court found that, “[a]lthough some medication abortions require a follow-up aspiration procedure, the number of those cases is exceedingly small and can generally be handled in an outpatient setting.” Abbott III, 2020 WL 1815587, at *3, 1 14 (citing Levison Decl. 1 9; Schutt-Aine Decl. 1 12). The Levinson paragraph cited speaks only to the frequency of hospitalization; it says nothing about how many medication abortions require follow-up aspiration. See App. 373. Nor does the cited Schutt- Aine paragraph provide any support for the frequency of follow-up aspiration. See App. 129. Schutt-Aine states that “[m]ajor complications—defined as complications requiring hospital admission, surgery or blood transfusion— occur in less than one-quarter of one percent (0.23%) of all abortion cases.” App. 129 (citing Ushma Upadhyay, et al., Incidence of Emergency Department Visits and Complications After Abortion, 125 Obstetrics & Gynecology, 175 (2015)). But Figure 1 of the cited article clarifies that subsequent uterine aspirations (i.e., surgical abortions) were not considered “surgery” within the meaning of the article. See Upadhyay, 125 Obstetrics & Gynecology at 176. Petitioners, by contrast, submitted evidence demonstrating the rate of medication abortions resulting in incomplete abortions, which are treated either with a repeat dose of medication or aspiration.[26] In our court, Respondents contend those numbers are outdated.[27] Analysis of such conflicting evidence is hard; it requires careful parsing. We reach no conclusions on the point. District courts, who can make fact findings after adversarial hearings, are better suited to the task. Here, however, the district court declined to avail itself of those tools, instead cancelling the scheduled preliminary injunction hearing and issuing a second TRO that adopted all 30 of Respondents’ proposed findings without citing or discussing a single declaration submitted by Petitioners. To be sure, a district court need not “recite every piece of evidence supporting its findings.” Schlesinger, 2 F.3d at 139. But “the record must nevertheless support the district court’s decision.” Sierra Club, Lone Star Chapter v. F.D.I.C., 992 F.2d 545, 551 (5th Cir. 1993). Here the record fails to do so. The failure to parse the evidence led the district court to reach legally erroneous results in two respects. First, under Whole Woman’s Health v. Hellerstedt, to determine whether a law “unduly burdens” the abortion right, a court must “consider the burdens a law imposes on abortion access together with the benefits those laws confer.” 136 S. Ct. 2292, 2309-10, 2319 (2016). The April 9 TRO does not meaningfully weigh either one. As noted, the order does not cite or discuss a single declaration submitted by Petitioners explaining the benefits of GA-09. Nor does the order articulate the burden of a delay or why that delay should be considered a “ban” on abortion. The record belies any such notion. Medication abortion is available until 10 weeks LMP, and surgical abortion until 22 weeks LMP. Given that GA-09 had only a 30-day duration, no woman would be pushed beyond the legal limit by a 30-day delay in obtaining a medication abortion. Moreover, health risks of a delay are mitigated because GA-09, by its terms, permits procedures that a patient’s physician determines are “immediately medically necessary to correct a serious medical condition of, or to preserve the life of, a patient who without immediate performance of the surgery or procedure would be at risk for serious adverse medical consequences or death.” Abbott II, 2020 WL 1685929, at *10 (quoting GA-09). The district court factored none of this into its cursory analysis. That weighing of burdens versus benefits would be inadequate under Hellerstedt in normal circumstances. A fortiori it is inadequate under the Jacobson framework, which asks whether burdens outweigh the benefits “beyond question.” 197 U.S. at 31. Moreover, as we have explained, the Supreme Court has approved “a wide variety of abortion regulations . . . that in practice can occasion real-world delays of several weeks.” Abbott II, 2020 WL 1685929 at *10 (quoting Garza v. Hargan, 874 F.3d 735, 755 (D.C. Cir. 2017) (en banc) (mem.) (Kavanaugh, J., dissenting)). That leads us to the second legal error resulting from the district court’s findings: they treat a medication abortion as an absolute right. But the constitutional right to abortion does not include the right to the abortion method of the woman’s (or the physician’s) choice. Gonzales, 550 U.S. at 164-65. On this record it was patently erroneous to find that a mere 30-day postponement of medication abortions “beyond question” violates Casey. d. The Pennhurst doctrine. We address an additional point that arose during our consideration of Petitioners’ emergency stay motion, because it may become important as the litigation continues. In the April 9 TRO, the district court adopted Respondents’ proposed fact finding that “[m]edication abortion is not a surgery or procedure.” Abbott III, 2020 WL 1815587 at *3, 1 10; cf. ECF 56-2, Plaintiffs Proposed Order 1 10 (“Medication abortion is not a surgery or procedure.”).[28] When considering Petitioners’ stay motion, we expressed uncertainty as to whether medication abortions were covered by GA-09, given ambiguity in the Texas Medical Board’s guidance on the order. See Abbott V, 2020 WL 1866010, at *3. For that reason, we denied a stay as to the part of the TRO applicable to medication abortions, while “express[ing] no ultimate decision on the ongoing mandamus proceeding.” Id. We have since benefitted from additional briefing on this issue. Given the lack of legal analysis in the April 9 order, we are unable to discern what impact the district court’s finding had on its decision to grant the TRO. Going forward, however, we caution that any relief ordering a state official to comply with state law would be barred by the Pennhurst doctrine. See Pennhurst State Sch. & Hosp. v. Halderman, 465 U.S. 89, 106 (1984). Under Pennhurst, a federal court may not grant “relief against state officials on the basis of state law.” Id. at 106. A federal court may determine state officials’ enforcement of state law violates a federal right, but it may not order state officials to conform their conduct to state law. See, e.g., Williams On Behalf of J.E. v. Reeves, 19-60069, 2020 WL 1638411, at *7 (5th Cir. Apr. 2, 2020) (under Pennhurst, “the rule announced in Ex parte Young cannot be used to redress a state official’s violation of state law”); Hughes v. Savell, 902 F.2d 376, 378 (5th Cir. 1990) (Pennhurst bars “a claim that state officials violated state law in carrying out their official responsibilities”). To the extent the April 9 TRO finds that GA-09 violates Casey by postponing medication abortions, we have already explained that it patently erred. But to the extent the TRO might be construed to order relief on a claim that state officials failed to conform their actions to state law, the TRO would violate Pennhurst. State health officials, who are Petitioners here, insist that GA-09′s postponement of “procedures” encompasses medication abortions. Pennhurst bars a federal court from considering a claim that those officials failed to comply with a proper interpretation of the state executive order. See, e.g., Hughes, 902 F.3d at 378 (quoting Pennhurst, 465 U.S. at 106) (explaining that “instruct[ing] state officials on how to conform their conduct to state law . . . conflicts directly with the principles of federalism that underlie the Eleventh Amendment”).[29] The district court should be aware of this issue in further proceedings. 2. The April 9 TRO Patently Erred by Exempting 18-Week Gestation from GA-09 We turn to the part of the April 9 TRO blocking application of GA-09 as to patients who “would reach 18 weeks LMP by April 21, 2020,” and who, in a physician’s judgment, are “unlikely to be able to obtain an abortion at an [ambulatory surgical center] before [her] pregnancy reaches the 22-week cutoff.” Abbott III, 2020 WL 1815587, at *6. For those patients, the district court concluded GA-09 would amount to “an absolute ban on abortion” that violates Casey. Id. Once again, the district court’s failure to apply the framework articulated in Abbott II led to a patently erroneous result that cannot be sustained on this record. As we explained in Abbott II, a state emergency measure like GA-09 violates the right to abortion if it “has no real or substantial relation” to the public crisis “or is, beyond all question, a plain, palpable invasion of [Casey].” 2020 WL 1685929, at *6 (quoting Jacobson, 197 U.S. at 31). Here, we take the district court’s conclusion to turn only on the second part of the analysis— whether GA-09 is “beyond all question” a violation of Casey to the extent it results in delaying a woman’s pregnancy to 18 weeks LMP. The district court’s treatment of GA-09 as “an absolute ban on abortion” as applied to this category of women was obviously wrong. Abbott III, 2020 WL 1815587, at *6. A woman who would be 18 weeks LMP when GA-09 expires has up to four weeks to legally procure an abortion in Texas. No case we know of calls that an “absolute ban” on abortion. Cf., e.g., Casey, 505 U.S. at 874 (explaining that “[n]umerous forms of state regulation might have the incidental effect of increasing the cost or decreasing the availability of medical care, whether for abortion or any other medical procedure”). The district court may have had in mind an as-applied challenge to GA- 09 on behalf of a woman facing this particular combination of circumstances. See, e.g., Gonzales, 550 U.S. at 167 (explaining that “as-applied challenges” are “the proper manner to protect the health of the woman if it can be shown in discrete and well-defined instances” that particular procedures are required). That would require evidence of “discrete and well-defined instances” sufficient to support such a challenge, id., but the district court cited none and we can find none in the record. Respondents attempt to bridge this gap by relying on a new affidavit from a hotline coordinator at an abortion-funding nonprofit. But that affidavit speaks only in general terms about women at later stages of pregnancy and does not even attempt to identify any “discrete and well-defined instances” of a woman in the 18-week category sufficient to support an as- applied challenge here. See App. 439-44. Respondents also speculate that, due to patient backlogs and the burden of traveling to one of the limited number of Texas ASCs, women in the 18-week category will not be able to obtain an abortion. Once again, this is the stuff of a possible as-applied challenge. But we know of no precedent saying that it violates Casey “beyond question” when a generally applicable emergency health measure causes backlogs and travel delays for women seeking abortion. In fact, even outside of a public health crisis, the Supreme Court has “recognize [d] that increased driving distances do not always constitute an ‘undue burden.’” Hellerstedt, 136 S. Ct. at 2313 (quoting Casey, 505 U.S. at 885-87). To the contrary, the Court has treated increased travel distance only as one factor that—”when taken together with others” such as “the virtual absence of any health benefit”—could support a conclusion of undue burden under Casey on a particular record. Id. (emphasis added). Perhaps in the context of a preliminary injunction hearing, Respondents will be able to adduce evidence to support an as-applied challenge to GA-09 (or its successor order, GA-15) along these lines. But the record presently before the district court fails to provide even an arguable basis to conclude that GA- 09, as applied to women in the 18-week category, is “beyond all question, a plain, palpable invasion of [Casey].” AbbottII, 2020 WL 1685929, at *6 (quoting Jacobson, 197 U.S. at 31). 3. The April 9 TRO Did Not Patently Err by Exempting 22-Week Gestation from GA-09. The district court also concluded that GA-09 “beyond question” violates Casey as applied to a woman who “would otherwise be denied access to abortion entirely because . . . [her] pregnancy would reach 22 weeks LMP” before GA-09 expires. Abbott III, 2020 WL 1815587, at *6. While we harbor some doubts about the evidentiary basis for the district court’s conclusion, we conclude that any error is not so clear and indisputable as to warrant mandamus. Unlike the 18-week category, Respondents have adduced some evidence that they have clients who will reach 22 weeks LMP during the operation of GA-09. See App. 103, 353, 442. While this evidence is secondhand, and thus weak, we cannot conclude it was a “clear abuse of discretion” for the district court to rely on it at this early stage. Abbott II, 2020 WL 1685929, at *4. The district court concluded that GA-09′s delay of non-essential medical procedures would operate as a permanent ban on abortion for women in this category, and that the order’s burdens far outweighed its benefits as to those women. Again, given the weak evidence, we are not fully satisfied with this cursory conclusion. Further, it remains unclear whether GA-09′s exception for “patient[s] who without immediate performance of the surgery or procedure would be at risk for serious adverse medical consequences . . . as determined by the patient’s physician,” id. at *3, already covers women in these circumstances. But Petitioners’ arguments do not convince us, at this early stage, that the district court’s order enjoining GA-09 as to women who will reach 22 weeks LMP during the order’s operation was so patently erroneous that mandamus is appropriate. Cf. Gee, 941 F.3d at 158 (noting that mandamus is only appropriate “for really extraordinary causes”). As a result, we conclude Petitioners have not shown entitlement to the writ of mandamus as to this part of the TRO. * * * To sum up, Petitioners have shown entitlement to the writ of mandamus as to the parts of the April 9 TRO that: • restrain enforcement of GA-09 as a “categorical ban on all abortions provided by plaintiffs”; restrain enforcement of GA-09 after 11:59 p.m. on April 21, 2020; restrain the Governor and Attorney General; restrain enforcement of GA-09 as to medication abortions; restrain enforcement of GA-09 as to abortions for patients who will reach 18 weeks LMP during the operation of GA-09 and would be “unlikely” to obtain abortion services in Texas. Petitioners have not demonstrated entitlement to the writ as to that part of the April 9 TRO that: restrains enforcement of GA-09 as to patients “who, based on the treating physician’s medical judgment, would be past the legal limit for an abortion in Texas—22 weeks LMP—on April 22, 2020.” IV. The other two requirements for mandamus relief are satisfied here. First, Petitioners “‘have no other adequate means’ to obtain the relief they seek.” Abbott II, 2020 WL 1685929, at *13. TROs, unlike preliminary injunctions, are not appealable. See Smith v. Grady, 411 F.2d 181, 186 (5th Cir. 1969); see also 28 U.S.C. § 1292. Although Petitioners argued in their separate appeal that the TRO at issue here has the “actual content, purport, and effect” of a preliminary injunction, Smith, 411 F.2d at 186, we concluded otherwise and dismissed that appeal for lack of jurisdiction. Second, for substantially the same reasons set out in Abbott II, “[w]e are persuaded that this petition presents an extraordinary case justifying issuance of the writ.” Abbott II, 2020 WL 1685929, at *15. As we stated there, the current global pandemic has caused a serious, widespread, rapidly-escalating public health crisis in Texas. Petitioners’ interest in protecting public health during such a time is at its zenith. In the unprecedented circumstances now facing our society, even a minor delay in fully implementing the state’s emergency measures could have major ramifications . . . . Id. The district’s failure to apply Jacobson and its usurpation of the state’s power by second-guessing “the wisdom and efficacy of [its] emergency measures” are just as extraordinary now as they were on April 7. Id. Moreover, the issues addressed in this litigation “have an importance beyond the immediate case.” Id. (quoting Volkswagen, 545 F.3d at 318). “[W]e are aware of nothing that would render the exercise of our discretion to issue the writ inappropriate.” Id. (quoting Volkswagen, 545 F.3d at 319). We therefore exercise our discretion to grant mandamus relief. CONCLUSION The petition for writ of mandamus is GRANTED IN PART and DENIED IN PART. The district court is directed to vacate any part of the April 9 TRO that (1) restrains enforcement of GA-09 as a “categorical ban on all abortions provided by Plaintiffs”; (2) restrains the Governor and Attorney General; (3) restrains enforcement of GA-09 after 11:59 p.m. on April 21, 2020; (4) restrains enforcement of GA-09 as to medication abortions; and (5) restrains enforcement of GA-09 as to abortions for patients who will reach 18 weeks LMP during the operation of GA-09 and would be “unlikely” to obtain abortion services in Texas. We do not grant the writ or direct vacatur as to that part of the April 9 TRO restraining enforcement of GA-09 as to patients “who, based on the treating physician’s medical judgment, would be past the legal limit for an abortion in Texas—22 weeks LMP—on April 22, 2020.” Any portions of our April 10 administrative stay remaining in effect are LIFTED. As indicated in Abbott II, any future appeals or mandamus petitions in this case will be directed to this panel and will be expedited. See Gee, 941 F.3d at 173; In re First South Sav. Ass’n, 820 F.2d 700, 716 (5th Cir. 1987). The mandate shall issue forthwith. JAMES L. DENNIS, Circuit Judge, dissenting in part. For the second time in as many weeks, the majority invokes the “drastic and extraordinary remed[y]” of mandamus, Ex parte Fahey, 332 U.S. 258, 259 (1947), simply to second guess the district court’s reasonable evaluation of the evidence and to interfere with its inherent power to control the proceedings before it. In so doing, the majority once again places us at odds with seemingly every other federal court to have considered whether the need to conserve hospital capacity and personal protective equipment (“PPE”) during the current COVID-19 pandemic can justify so drastically curtailing the constitutional right to an abortion. See In re Abbott, No. 20-50264, 2020 WL 1685929, at *16 (5th Cir. Apr. 7, 2020) (Abbott II) (Dennis, J., dissenting) (collecting cases).[30] This second ruling is particularly inappropriate because, although the district court properly fulfilled this court’s previous mandate— unwarranted though the mandate may have been—the majority now moves the goal posts and chastises the district court for not abiding by a series of phantom instructions that can be found nowhere in its previous order. At bottom, the majority simply disagrees with the district court’s decisions on matters that are squarely within its discretion. This is not a proper use of “one of the most potent weapons in the judicial arsenal.” In re JPMorgan Chase & Co., 916 F.3d 494, 504 (5th Cir. 2019) (quoting Cheney v. U.S. Dist. Court for D.C., 542 U.S. 367, 380 (2004)). This Circuit thus once again does not apply the applicable rules of law because of the subject matter of the case, and, equally troubling, ignores the words of its own ruling from less than two weeks ago. I again echo the words of a colleague in dissent in a case now before the United States Supreme Court: “It is apparent that when abortion comes on stage it shadows the role of settled judicial rules.” June Med. Services L.L.C. v. Gee, 905 F.3d 787, 816 (5th Cir. 2018) (Higginbotham, J., dissenting), cert. granted, 140 S. Ct. 35, 204 L. Ed. 2d 1193 (2019). I. The facts and procedural history of this case have been documented in detail in our previous decision. See In re Abbott, No. 20-50264, 2020 WL 1685929 at *1-4 (5th Cir. Apr. 7, 2020) (Abbott II); id. at *17 (Dennis, J., dissenting). To briefly recount, on March 22, 2020, the Governor of Texas issued executive order GA-09 in response to the current COVID-19 pandemic and the accompanying shortage of personal protective equipment (“PPE”) and hospital capacity. GA-09 requires all Texas healthcare providers to postpone all surgeries and procedures that are not immediately medically necessary to correct a serious medical condition of, or to preserve the life of, a patient who without immediate performance of the surgery or procedure would be at risk for serious adverse medical consequences or death, as determined by the patient’s physician. Tex. Exec. Order No. GA-09 (Mar. 22, 2020), https://gov.texas.gov/uploads/files/press/EO-GA_09_COVID-19_hospital_capacity_IMAGE_03-22-2020.pdf. The order contains an exception for “any procedure that, if performed in accordance with the commonly accepted standard of clinical practice, would not deplete the hospital capacity or the [PPE] needed to cope with the COVID-19 disaster.” Id. Violations of GA-09 are punishable by criminal penalties and, by virtue of a subsequent emergency rule with identical requirements that was issued by the Texas Medical Board, see 22 Tex. Admin. Code § 187.57, may effectively result in the suspension or restriction of a practitioner’s license.[31] By its terms, GA-09 remains in effect until 11:59 p.m. on April 21, 2020. The day after GA-09 was issued, the Texas Attorney General released a press release stating, among other things, that GA-09 applied to “any kind of abortion that is not medically necessary to preserve the life and health of the mother,” and that “[t]hose who violate the governor’s order will be met with the full force of the law.” App. at 31.[32] The Respondents, who provide abortion services in Texas, filed suit in district court under 42 U.S.C. § 1983 against the Petitioners, who are various state officials. The Respondents asserted that the application of GA-09 to prohibit abortion violated, inter alia, substantive due process. The Respondents sought to enjoin the Petitioners from enforcing GA-09 as applied to abortion, and, after reviewing argument and evidence on the point from both parties, the district court issued a temporary restraining order (“TRO”) doing just that. See Planned Parenthood Ctr. for Choice v. Abbott, No. A-20-CV-323-LY, 2020 WL 1502102 (W.D. Tex. Mar. 30, 2020). Before the district court could hold a scheduled hearing on whether to issue a longer preliminary injunction, the Petitioners filed a petition for mandamus with this court, and on April 7, the majority granted the petition and ordered that the district court vacate its TRO. Abbott II, 2020 WL 1685929 at *16. The Respondents then moved for a second, more limited TRO in the district court, which the court granted. Planned Parenthood Ctr. for Choice v. Abbott, No. A-20-CV-323-LY, 2020 WL 1815587, at *7 (W.D. Tex. Apr. 9, 2020) (Abbott III). The second TRO restrained the Petitioners from enforcing GA-09 (1) “as a categorical ban on all abortions provided by [the Respondents]“; (2) as a prohibition on “medication abortions”; (3) as a prohibition on “procedural abortion[s] [for] any patient who, based on the treating physician’s medical judgment, would be more than 18 weeks LMP[[33]] on April 22, 2020, and [who are] likely unable to reach an ambulatory surgical center in Texas or to obtain abortion care” (“the 18-week category”); and (4) as a prohibition on “procedural abortion[s] [for] any patient who, based on the treating physician’s medical judgment, would be past the legal limit for an abortion in Texas—22 weeks LMP—on April 22, 2020″ (“the 22-week category”). Id. The Petitioners filed a second petition for a writ of mandamus with this court the following day. II. Petitioners once again ask that we direct the district court to vacate its TRO. As noted, mandamus is an “extraordinary remedy” that should only issue in “exceptional circumstances.” See Cheney, 542 U.S. at 380 (quoting Will v. United States, 389 U.S. 90, 95 (1967)); In re Lloyd’s Register N. Am., Inc., 780 F.3d 283, 288, 294 (5th Cir. 2015); In re Volkswagen of Am., Inc., 545 F.3d 304, 309, 311 (5th Cir. 2008). It is not sufficient for the petitioners to prove simply “that the court erred.” In re Occidental Petroleum Corp., 217 F.3d 293, 295 (5th Cir. 2000). Rather, mandamus relief generally requires that (1) “the party seeking issuance of the writ [must] have no other adequate means to attain the relief he desires—a condition designed to ensure that the writ will not be used as a substitute for the regular appeals process”; (2) “the petitioner[s] must satisfy the burden of showing that [their] right to issuance of the writ is clear and indisputable”; and (3) “the issuing court, in the exercise of its discretion, must be satisfied that the writ is appropriate under the circumstances.” Cheney, 542 U.S. at 380-81 (internal quotation marks and citations omitted). The “clear and indisputable” prong of this test is not met here. That the Petitioners’ right to relief is indeed disputable should be evident by the very existence of this dissent and the many other courts that have concluded that relief is not warranted in very similar circumstances. But I will elaborate. To establish a clear and indisputable right to relief, the Petitioners must show that the district court not only committed a “clear abuse[] of discretion,” but also that the abuse “produce[d] patently erroneous results.” In re Lloyd’s Register N. Am., Inc., 780 F.3d at 290 (quoting In re Volkswagen of Am., Inc., 545 F.3d at 312). Neither precondition is met here. A. The majority asserts that the district court abused its discretion in several ways. None of these assertions warrants mandamus relief. 1. The majority first engages in what would appear to be an academic exercise, concluding that the district court erred by restraining the Petitioners’ conduct in ways that apparently had no practical impact even under the majority’s reasoning. Mandamus relief cannot be warranted to fix a mistake that is of no consequence. To start, the majority faults the district court for restraining the Petitioners from enforcing GA-09 “as a categorical ban on all abortions” because it does not interpret GA-09 to be a categorical ban on all abortions. Majority at 9 (citing Abbott III, 2020 WL 1815587, at *7). As the majority acknowledges, however, a federal court may determine that a state official’s purported enforcement of state law would violate a federal right, and this principle applies regardless of whether that enforcement is a correct interpretation of the state law. Majority at 31; See Louise B. v. Coluatti, 606 F.2d 392, 399 (3d Cir. 1979) (“To put the matter more bluntly, where a state violates federal law, it is no better off because it also violates its own law.”). Whether GA-09 actually is a categorical ban on abortions under state law is therefore irrelevant. If the district court correctly found a credible threat that the Petitioners would enforce GA-09 as a categorical ban on abortion, that factual finding is sufficient to restrain such enforcement as a violation of a federal constitutional right. See Susan B. Anthony List v. Driehaus, 573 U.S. 149, 159 (2014) (stating that when “plaintiffs face a credible threat of enforcement,” they “should not be required to await and undergo a criminal prosecution as the sole means of seeking relief” (internal quotations omitted)). But as I said, this discussion is academic. By restraining the Petitioners from enforcing GA-09 as to medication abortions and those abortions falling into the 18-week and 22-week categories, the district court’s TRO also necessarily prevented the Petitioners from enforcing GA-09 as a categorical ban on abortions. Neither the Petitioners nor the majority identify any possible conduct that would violate the “categorical ban” prohibition in the district court’s order that would not also violate the prohibition on enforcing GA-09 against providers of medication abortions or those that provide abortions that fall into the 18-week or 22-week categories. Quite simply, the TRO’s prohibition on enforcing GA-09 “as a categorical ban on all abortions” has no independent effect; it does not burden the Petitioners because it does not alter the Petitioners’ rights or responsibilities in any way. Accordingly, even if correct regarding the error, the majority’s decision to order this portion of the order vacated also does not change the rights and responsibilities of any party and serves no purpose but to point out the district court’s purported mistake. Invoking “one of the most potent weapons in the judicial arsenal,” In re JPMorgan Chase, 916 F.3d at 504, simply to correct an alleged legal error with no practical consequences is inappropriate. Indeed, we have explicitly stated that mandamus is not warranted upon a showing “merely that the court erred.” Occidental Petroleum Corp., 217 F.3d at 295. The Petitioners and the majority have not demonstrated how the district court’s purported error on this point could produce “patently erroneous results,” In re Volkswagen of Am., Inc., 545 F.3d at 312, when, as a practical matter, it does not produce any results at all. Similarly, the majority concludes that, because the district court restrained the enforcement of GA-09 past its nominal April 22 expiration date, the district court abused its discretion by not narrowly tailoring its TRO to end when the executive order potentially expired. Majority at 9-10. But the majority fully acknowledges that, following GA-09′s expiration, there will be “no enforcement of GA-09 for a court to restrain,” Majority at 10, and so the district court’s TRO will have no effect on the Petitioners after the executive order expires. By contrast, under the majority’s reasoning, had GA-09 been extended, the district court would have had to again extend its TRO in order to maintain the status quo until the scheduled preliminary injunction hearing—something that it is not even clear the district court would have authority to do under FED. R. CIV. P. 65(b)(2), which some authorities have held limits a district court to one TRO extension.[34] See Clements Wire & Mfg. Co. v. N. L. R. B, 589 F.2d 894, 896 (5th Cir. 1979) (stating that TROs “must expire not later than 20 days after issuance,” when Rule 65(b) imposed a 10-day time limit); U.S. Dep’t of Labor v. Wolf Run Mining Co., 452 F.3d 275, 281 (4th Cir. 2006) (stating “a TRO is limited in duration to 10 days plus one 10-day extension” under the former time limit (emphasis added)). The extension of the TRO was a reasonable action by the district court that was well within its discretion—either its TRO would be harmless and would not affect any rights or responsibilities after GA-09′s expiration, or it would be appropriately tailored to its purpose. We now know that, fortunately, GA-09 will indeed expire on April 22,[35] and the majority’s order will not wreak the harm that it might have had GA-09 been extended and the district court been left with a gap between the expiration of the TRO and the preliminary injunction hearing that it may arguably have been powerless to fill. But again, the majority utilizes mandamus, a “‘drastic and extraordinary’ remedy ‘reserved for really extraordinary causes,’” Cheney, 542 U.S. at 380 (quoting Ex parte Fahey, 332 U.S. at 259-60), simply to correct what it wrongly perceives to be a run-of-the- mill legal error that the majority acknowledges has no practical consequences. The majority next concludes that the district court erred by failing to dismiss the Texas Governor and Attorney General and by restraining them from enforcing GA-09 in the proscribed manner because neither officer has the “connection” to enforcement of GA-09 needed to overcome sovereign immunity under Ex parte Young, 209 U.S. 123 (1908). Majority at 10-13. With respect to the Attorney General, the majority concludes that neither the fact that Texas law permits the Attorney General to participate in prosecutions for violations of GA-09, see Tex. Gov’t CODE § 402.028, nor that the Attorney General publicly singled out abortion providers for potential enforcement; stated that GA-09 prohibits all abortion except in the case of a medical emergency; and threatened that “[t]hose who violate the governor’s order will be met with the full force of the law,” App. at 30, are sufficient to establish a connection to GA-09 enforcement. Our cases have not explicitly held as much, and indeed we have previously stated that authority and a willingness to enforce a law can be inferred by an official’s threats to do so. City of Austin v. Paxton, 943 F.3d 993, 1001 (5th Cir. 2019) (“[T]he fact that Paxton sent letters threatening enforcement of the DTPA makes it clear that he had not only the authority to enforce the DTPA, but was also constraining the manufacturer’s activities, in that it faced possible prosecution if it continued to make and distribute its products.”). It is difficult to see, then, how the district court’s preliminary[36] determination that the Attorney General had a connection to enforcement was so contrary to established law as to constitute a “clear abuse of discretion,” In re Volkswagen of Am., Inc., 545 F.3d at 312, and thereby justify this court’s extraordinary intervention on this point. But even assuming arguendo that the majority is correct, the sum effect of restraining the Governor and Attorney General from enforcing GA-09 would be nil if they lack any authority to enforce GA-09 in the first place. Thus, the majority acts again to use the “drastic and extraordinary remedy” of mandamus, where it at most has little, if any, practical effect. Fahey, 332 U.S. at 259. I therefore disagree that mandamus is appropriate with respect to any of these alleged errors. 2. Next the majority concludes that the district court violated the “mandate rule” by not following the instructions issued in Abbott II. Majority at 14 (citing United States v. Smith, 814 F.3d 268, 273 (5th Cir. 2016)). But a review of the district court’s substantive and procedural decisions makes clear that the court complied fully with the majority’s previous directives. a. The majority first argues that the district court failed to apply the legal framework that the majority previously derived from Jacobson v. Commonwealth of Massachusetts, 197 U.S. 11 (1905), and Planned Parenthood v. Casey, 505 U.S. 833 (1992). Majority at 13. This is contradicted by the district court’s plain statement that applying GA-09 to the classes of abortions identified in its TRO “violates the standards set forth in both [Casey] and [Jacobson].” Abbott III, 2020 WL 1815587. The majority appears to regard this as only a formalistic recitation. On the contrary, the district courts’ findings and conclusions makes clear that it properly considered each step of the majority’s framework. Specifically, the test the majority previously formulated required the district court to determine (1) whether applying GA-09 to the abortions at issue[37] lacks a “real or substantial relation” to the current public health crisis, and (2) whether the benefits of applying GA-09 to these abortions are “beyond all question” outweighed by its burden on the constitutional right to an abortion, thus creating an undue burden under Casey. Abbott II, 2020 WL 1685929, at *7, *9. The district court faithfully made findings and conclusions tied to both steps of this analysis. With respect to the first, the majority stated in its previous opinion that the district court should “inquire whether Texas has exploited the present crisis as a pretext to target abortion providers sub silentio.” Id. at *13 (citing Lawton v. Steele, 152 U.S. 133, 137 (1894)). On this point, the district court found that “[m]edication abortion is not a surgery or procedure” within conventional definitions of the terms and that “[p]roviding medication abortion does not require the use of any PPE,” which would suggest that it is not covered by the plain language of the prohibition within GA-09. Abbott III, 2020 WL 1815587, at *3. And the court found that “[p]hysicians are continuing to provide [other] obstetrical and gynecological procedures comparable to abortion in PPE use or time-sensitivity, based on their professional medical judgment” without hinderance from GA-09. Id. at *5. And, though this significantly overlaps with the second prong of the majority’s test, the district court made extensive findings as to why applying GA-09 to abortions “will not conserve PPE” and “will not conserve hospital resources”—the two explicitly stated goals of the executive order. Id. at *4. These findings all suggest that the application of GA-09 to abortions—particularly medication abortions—is pretextual and not motivated by any desire to conserve PPE or hospital capacity, meaning it lacks a “real or substantial relation” to the current public health crisis under the framework the majority instructed the district court to employ. Abbott II, 2020 WL 1685929, at *7. The district court’s analysis is even more explicit at the second stage of the inquiry in which the majority instructed it to weigh the benefits of applying GA-09 to abortions against the burden on the constitutional right to an abortion. Id. at *9. As stated, the district court made a range of findings indicating that applying GA-09 to abortions not only would not result in a public health benefit but in fact would be a net drain on PPE and hospital capacity and would be otherwise harmful to public health. The court made specific findings as to how much and what kind of PPE are consumed during the performance of medication and procedural abortions and their surrounding services, as well as the frequency with which complications require hospitalization with each method. Abbott III, 2020 WL 1815587, at *3-4. The court then compared this figure to the amount of each resource typically consumed by a woman continuing a pregnancy and found that substantially more of each resource is consumed “at each stage of the [continued] pregnancy” than in a pre-viability abortion. [38] Id. at *4. And the district court noted that continuing a pregnancy requires significantly more “in-person healthcare” and that many women denied an abortion by GA-09 are traveling out-of-state to obtain one, increasing their risk of contracting COVID-19. Id. at *4-5. The district court then made a variety of findings regarding the burden that applying GA-09 to the classes of abortion at issue placed on the constitutional right to abortion. The court found that the Respondents had “turned away hundreds of patients seeking abortion care,” and that “[t]here will be significant pent-up need for abortion care when the Executive Order expires,” resulting in further delays. Id. at *5. The court noted that progressively more invasive techniques are required to perform an abortion the longer that it is delayed. Id. And the court found that the health risks, financial costs, and emotional cost of an abortion increases with gestational age, meaning the delays make abortions riskier and more cost prohibitive. Id. at *5-6. The court also found that for many women, the delay would result in them being effectively denied a legal right to an abortion in Texas, either because they would exceed the 22-week maximum legal limit or they would reach a gestational age at which they were legally required to go to a facility that they did not have access to in order to receive an abortion. Id. The district court then weighed these benefits and burdens against each other, just as the majority instructed, and concluded that “based on the court’s findings of fact, it is beyond question that the Executive Order’s burdens outweigh the order’s benefits as applied” to the classes of abortion at issue. Id. at *6. And in doing so, it explicitly relied Jacobson. It is one thing for the majority to disagree with the district court’s findings or its weighing of the relative benefits and burdens of applying GA-09 to these abortions (though, as will be discussed, the majority is wrong to do so). But it is simply inaccurate for the majority to conclude that the district court disregarded its instructions to apply a legal framework that it is abundantly clear the district court in fact applied. I therefore disagree that the district court violated the mandate rule in this respect and, therefore, that mandamus is appropriate to correct the supposed error. b. The majority next reprimands the district court for entering a TRO at all. Majority at 14-19. The majority seems to have wanted the district court to instead proceed directly to an adversarial hearing on a preliminary injunction without issuing another TRO to preserve the status quo until that hearing could be held. This requirement is stated nowhere in the previous mandamus order and cannot be reconciled with the principle that district courts have broad discretion in ordering their affairs. To be sure, the majority made several references to the sort of evidence that could be adduced and arguments that could be made at an upcoming preliminary injunction hearing. See, e.g., Abbott II, 2020 WL 1685929, at *2 (“The district court has scheduled a telephonic preliminary injunction hearing for April 13, 2020, when all parties will presumably have the chance to present evidence on the validity of applying GA-09 in specific circumstances.”); id. at *9 (“Respondents will have the opportunity to show at the upcoming preliminary injunction hearing that certain applications of GA-09 may” violate Casey); id. at *12 (“These are issues that parties may pursue at the preliminary injunction stage[.]“). But even assuming arguendo these off-hand comments could be construed as a directive to hold a preliminary injunction hearing—they obviously cannot—at no point did the majority suggest the district court was not permitted to issue a TRO to prevent irreparable harm in the interim until a hearing could be held. Nor is it clear by what authority the majority would have imposed such a restriction if it had been contained in the previous mandamus order. It is well established that “[d]istrict courts have broad discretion to evaluate the irreparability of alleged harm and to make determinations regarding the propriety of injunctive relief.” Wagner v. Taylor, 836 F.2d 566, 575-76 (D.C. Cir. 1987). This broad discretion extends to “manag[ing] the timing and process for entry of all interlocutory injunctions—both TROs and preliminary injunctions[.]“ Ciena Corp. v. Jarrard, 203 F.3d 312, 319 (4th Cir. 2000). The majority and the Petitioners make much of the fact that the district court did not permit the Petitioners to “respond [to the second TRO motion] in writing” before entering the second TRO. Transcript of 4/9/20 Tele. Conf. at 14:39; see Majority at 18 (“Moreover, we called for additional evidence from both sides.”). But as the majority acknowledges, the district court was not required to do so, as the Rules explicitly allow a TRO to be entered ex parte. Majority at 14 (citing Fed. R. Civ. P. 65(b)(1)). Further, the Petitioners had already been permitted to make argument and introduce evidence in response to the Respondents’ first motion for a TRO. See Planned Parenthood Center for Choice v. Abbott, No. 1:20-cv-00323-LY (W.D. Tex.) (Dkt. No. 30) (March 30, 2020). It was not an abuse of the district court’s discretion for it to conclude that, given the time constraints and the potential for irreparable harm that had been established, it was imprudent to wait for further argument or evidence from the Petitioners before entering a TRO to preserve the status quo until the preliminary injunction hearing. All that was required was that “the opposing party [be] given a reasonable opportunity, commensurate with the scarcity of time under the circumstances, to prepare a defense and advance reasons why the [preliminary] injunction should not issue.” Ciena Corp., 203 F.3d at 319. The Petitioners would have had that opportunity at the rescheduled preliminary injunction hearing, and they could have filed a motion to dissolve the TRO with accompanying exhibits at any time under Rule 65(b)(4) if they truly believed that time was of the essence such that they could not wait for the hearing.[39] Instead, they filed the present motion for mandamus. The majority alternately argues that the district court defied the mandate by entering a TRO on essentially the same record that the previous mandamus order found was inadequate to support a TRO. Majority at 14-19. This is simply inaccurate. Following the district court’s first March 30 TRO order, which was based on ten declarations submitted by the Respondents, the Respondents filed nine additional declarations as supplements to their motion for a preliminary injunction. See Abbott, No. 1:20-cv-00323-LY (Dkt. No. 49) (April 2, 2020). And the Respondents included a new tenth declaration with their second motion for a TRO. See Abbott, No. 1:20-cv-00323-LY (Dkt. No. 56) (April 8, 2020). The district court thus had twice the number of declarations filed by the Respondents before it when it entered its second TRO than when it entered its first. In an attempt to overcome the fact that the record was clearly substantially more developed when the district court entered its second TRO, the majority argues that the Respondents cited to the supplemental declarations in their opposition to the Petitioners last mandamus petition, and so these declarations were included in the record that the majority reviewed and declared insufficient. Majority at 17-18. But the majority’s last mandamus order commented only on “the record before us.” Abbott II, 2020 WL 1685929, at *9. The fact that the Respondents cited to additional declarations in their briefing is of no moment. The additional declarations were not included in the appendix filed with the previous mandamus petition and they were not before the district court when it made the decision we reviewed. Further, the majority did not explicitly take judicial notice of the additional declarations, and the contents of the declarations are not the type of indisputable information suitable for judicial notice in any event. See Gov’t of Canal Zone v. Burjan, 596 F.2d 690, 694 (5th Cir. 1979) (noting that Federal Rule of Evidence 201, which permits judicial notice only of generally known facts or those that can be readily determined from an indisputably reliable source, applies at every stage of a proceeding). There was no reason for the district court—or anyone, for that matter—to assume the supplemental declarations were included in the majority’s review of “the record before us” that resulted in its conclusion that the evidence was inadequate. Abbott II, 2020 WL 1685929, at *9. Indeed, if the additional declarations were included in that review, the majority’s statement was arguably an improper advisory opinion on matters not before the court because the sufficiency of those declarations had no bearing on whether the district court’s previous decision was correct at the time it was made. Cf. United States v. Marine Shale Processors, 81 F.3d 1329, 1352 (5th Cir. 1996) (declining to give an advisory opinion on matters that did not affect the issues before it and that could be litigated in a later proceeding); Henry v. Dep’t of Hous. & Urban Dev. (HUD), Washington, D. C., 451 F.2d 355, 356 (8th Cir. 1971) (noting that the court could not give an advisory opinion as to the validity of a different complaint that was not the complaint the district court dismissed). In sum, the district court considered new evidence just as the majority instructed, and it applied the legal framework that the majority prescribed. The majority is thus wrong to now hold that the district court failed to fulfill the majority’s previous mandate simply because it did not meet new additional requirements that were not stated in the majority’s order, and I therefore disagree that this supposed defiance warrants mandamus. B. That the district court did not commit a clear abuse of discretion is sufficient reason on its own to deny mandamus. See In re Lloyd’s Register N. Am., Inc., 780 F.3d at 290 (quoting In re Volkswagen of Am., Inc., 545 F.3d at 312). But mandamus is also not warranted because, contrary to the majority’s contention, the district court’s analysis did not lead to patently erroneous results. Assuming arguendo that the majority’s interpretation of Jacobson is correct and the standard it articulated applies here, restraining the enforcement of GA-09 as a prohibition on the classes of abortion at issue was appropriate if (1) that enforcement is “pretextual—that is, arbitrary or oppressive” because it “has no real or substantial connection” to protecting public health during the COVID-19 epidemic; or (2) if that enforcement is “‘beyond all question, a plain, palpable invasion’ of the [constitutional] right to abortion.” Abbott II, 2020 WL 1685929, at *7, *8 (quoting Jacobson, 197 U.S. at 31). There is ample evidence in the record to support the conclusion that both of these requirements are met with respect to the applications of GA-09 the district court restrained through its second TRO, and the majority’s disagreement with the district court’s reasonable evaluation of this evidence does not make the TRO palpably erroneous. As an initial matter, the majority breezes past the pretext prong of its test, barely even mentioning it in its analysis of the results of the district court’s reasoning. This is likely because, in its previous order, the majority without explanation elected to analyze this question as whether GA-09 as a whole had a real or substantial relation to the current public health crisis, and so was able to easily conclude that it did. See id. at *8 (“The answer is obvious: the district court itself conceded that GA-09 is a valid emergency response to the COVID-19 pandemic.”). But this is the wrong focus of the inquiry. The Respondents were not challenging GA-09 as a whole, nor had the district court enjoined all applications of GA-09. The Respondents had challenged, and the district court had restrained, the application of GA-09 to abortions, and Jacobson itself makes clear that the question should not have been whether “GA-09 is a pretext for targeting abortion.” Id. at *13. It should have been whether the enforcement of GA-09, a concededly valid public health measure, was being used as a pretext to target abortions by state actors motivated by hostility to abortion rights. See Jacobson, 197 U.S. at 28 (“We say necessities of the case, because it might be that an acknowledged power of a local community to protect itself against an epidemic threatening the safety of all might be exercised in particular circumstances and in reference to particular persons in such an arbitrary, unreasonable manner, or might go so far beyond what was reasonably required for the safety of the public, as to authorize or compel the courts to interfere for the protection of such persons.” (emphases added)). Indeed, to hold otherwise would sanction the use of valid public health measures as tools for the arbitrary deprivation of any number of constitutional rights, regardless of the logical efficacy of applying the measure in that manner or even whether the enforcement was in fact motivated by a desire to further public health. See, e.g., On Fire Christian Ctr., Inc. v. Fischer, No. 3:20-CV- 264-JRW, 2020 WL 1820249, at *2 (W.D. Ky. Apr. 11, 2020) (finding that a ban on public gatherings during the pandemic, which would obviously be a valid public heath measure, was unconstitutional when used to prevent a drive- through Easter Sunday church service in which parishioners remained in their cars and had no direct personal contact (quoting Jacobson, 197 U.S. 11, 31 (1905)). Considered in this light, there was sufficient evidence in the record to conclude that the enforcement of GA-09 as a prohibition on all three of the classes of abortion at issue was pretextual and motivated not by a desire to advance public health, but rather to reduce the number of abortions performed for its own sake. To begin with, as the district court determined and as the majority acknowledged in a previous order in this case, see In re Abbott, No. 20-50296, 2020 WL 1866010 (5th Cir. Apr. 13, 2020), medication abortion, which in itself consists entirely of providing a patient with two sets of oral medication, is not a “surgery or procedure” under either the conventional definitions of those terms or the meaning assigned to them in informal guidance from the Texas Medical Board. See TEXAS MEDICAL BOARD, Frequently Asked Questions (FAQs) Regarding Non-Urgent, Elective Surgeries and Procedures During Texas Disaster Declaration for COVID-19 Pandemic (Mar. 29, 2020), https//www.tmb.state.tx.us/idl/59C97062-84FA-BB86-91BF- F9221E4DEF17 (last visited Apr. 19, 2020). And, as a wide range of declarations in the record establish, medication abortion consumes no PPE whatsoever when considered in isolation without the preceding ultrasound or post-abortion tests that Texas law requires. See App. at 73, 86, 91-92, 100, 110, 117, 134, 157. Medication abortion therefore does not appear to fall within the facial, plain meaning of GA-09′s prohibition on non-urgent elective “surgery or procedures,” and if it did, it would seem to fall into the exception for “any procedure that, if performed in accordance with the commonly accepted standard of clinical practice, would not deplete the hospital capacity or the personal protective equipment needed to cope with the COVID-19 disaster.”[40] Texas Executive Order No. GA-09 (Mar. 22, 2020), https://gov.texas.gov/uploads/files/press/E0-GA_09_C0VID-19_hospital_ capacity_IMAGE_03-22-2020.pdf. The Petitioners’ stated desire to nonetheless enforce the order against the providers of medication abortion raises a strong inference that the enforcement against abortion providers more generally is pretextual and motivated by hostility to abortion rights. Further supporting this inference are multiple declarations from practicing physicians stating that GA-09 is not being enforced to prohibit many obstetrical and gynecological procedures that consume as much or more PPE and hospital capacity than the categories of abortion at issue here, suggesting that abortions were being singled out for differential treatment. See App. at 368, 373-74. And, as discussed in more detail below, there is substantial evidence in the record that enforcing GA-09 against these categories of abortion leads to a net loss of PPE and hospital capacity because the amount of each resource consumed at every stage of a continued pregnancy is greater than the amount consumed in the performance of a medication or procedural abortion. See App. at 135, 372-74, 414. Based on this evidence, the district court could reasonably conclude at the TRO stage that Petitioner’s enforcement of GA-09 as a prohibition against the categories of abortion at issue here was “pretextual” and had “no real or substantial connection” to protecting public health during the COVID-19 epidemic. Abbott II, 2020 WL 1685929, at *7, *8 (quoting Jacobson, 197 U.S. at 31). Under the majority’s framework, this fact alone is enough to demonstrate that the district court’s determinations did not produce “patently erroneous results” as required for the issuance of mandamus. See In re Lloyd’s Register N. Am., Inc., 780 F.3d at 290 (quoting In re Volkswagen of Am., Inc., 545 F.3d at 312). But the majority doubly errs because there was also ample evidence to conclude that enforcement of GA-09 against each of these categories of abortion fails the second prong of its test by being “‘beyond all question, a plain, palpable invasion’ of the [constitutional] right to abortion.” Abbott II, 2020 WL 1685929, at *7, *8 (quoting Jacobson, 197 U.S. at 31). I will consider the two classes of abortions the majority vacates the TRO with respect to in turn. 1. The majority first considers whether the evidence was sufficient for a district court to reasonably conclude that enforcing GA-09 against providers of medication abortion “beyond all question” violated the constitutional right to an abortion. Majority at 19-31. The majority begins by asserting that the district court considered only the relative consumption of PPE associated with medication abortions under normal circumstances and asserts that there is no evidence documenting PPE usage rates during the current pandemic. Majority at 20-23. First, it is worth reiterating that the majority’s previous order did not include a directive to the district court to specify that findings it was making regarding relative usage were about rates during the current pandemic. The majority now changes the rules in order to find error where there is none. Its new requirement is based solely on the majority’s own supposition that, during the current pandemic, there is PPE used during medication abortions that would not otherwise be used by abortion providers furnishing other healthcare services, and that this increase shifts the balance between the relative benefits and burdens of applying GA-09 to prohibit medication abortions. There is no evidence for the majority’s supposition. In support of its contention that the rate of PPE usage has likely changed, the majority points to a declaration by an infectious disease expert that states “[n]ot wearing face masks and other PPE when caring for patients who are not under investigation for COVID 19 . . . exposes health care workers to transmission of infection” from asymptomatic patients. Majority at 22. But there is no indication that the abortion providers would not wear the same amount of PPE “caring for patients” in ways other than providing abortion. The majority fails to make the simple logical inference that, if medication abortion requires no PPE under normal conditions, it requires no more PPE than would be used by medical staff providing other services under pandemic conditions. For the majority’s premise to be correct, one would have to assume that abortion providers only (or at least primarily) provide abortion services and would not fill canceled abortion appointment slots with appointments for other medical services that would bring them into personal contact with patients at a similar frequency.[41] Even assuming arguendo that medication abortions do consume more PPE in a pandemic, the pertinent question is not whether prohibiting medication abortion prevents the use of some marginal amount of PPE. It is whether it creates a net benefit that outweighs its burden on the constitutional right to abortion. See Whole Woman’s Health v. Hellerstedt, 136 S. Ct. 2292, 2310 (2016). As I stated before, there was no evidence in the record suggesting that any PPE purchased by or in the possession of abortion providers that would be conserved by applying GA-09 to abortions could be redirected to the COVID-19 response, nor have the majority or Petitioners articulated any logical way in which this could be so. Abbott II, 2020 WL 1685929, at *22 (Dennis, J., dissenting). Moreover, there are multiple declarations in the record from health care professionals documenting that any increase in PPE consumption from medication abortion during the pandemic is more than matched by an increase in PPE consumption from the necessary medical services associated with continuing a pregnancy. See, e.g., App. at 135 (“By comparison, even if a provider of prenatal care reduces the scheduling of such care during the COVID-19 outbreak, it will still involve use of masks, sterile gloves, and potentially other PPE during multiple visits. A patient continuing pregnancy will thus require significantly more PPE than a patient presenting for abortion.”); App. at 375 (stating that “most prenatal and postpartum care is continuing” during the COVID-19 pandemic and cannot be done remotely before concluding that “requiring people to continue unwanted pregnancies utilizes more PPE and more hospital resources than abortion care”). Thus, there is more than sufficient evidentiary support to conclude that applying COVID-19 to prohibit medication abortions does not preserve PPE during the current pandemic. The majority attempts to overcome this basic conclusion by asserting that the state made a policy judgment that it was more important to conserve PPE in the near term than the long term in order to “flatten the curve.” Majority at 25. But this ignores the evidence that continuing a pregnancy results in more PPE usage at every stage of the pregnancy than is typically used in an abortion of any sort. See App. at 414 (“[T]he imaging and laboratory tests alone needed during early pregnancy require the use of more PPE than is typically used in connection with an abortion.”) There is thus ample evidentiary support to conclude that applying GA-09 to medication abortions results in no conservation of PPE in the short or long term. Similar evidence exists with respect to hospital capacity; when a pregnancy is continued, more hospital beds and resources are consumed than when a woman obtains a pre-viability abortion, and there is no indication that the current pandemic has changed this. See, e.g., App. at 375. And significant evidence supports a conclusion that women who are unable to obtain an abortion because of GA-09 will travel out of state to obtain one, increasing their risk of contracting COVID-19 and spreading it to others.[42] See, e.g., App. at 258-59, 311. Thus, this is not an instance in which the district court made a “choice . . . between two reasonable responses to a public crisis,” that should have been “left to the governing state authorities.” Abbott II, 2020 WL 1685929, at *12 (citing Jacobson, 197 U.S. at 30). This is an instance in which the district court logically concluded that applying GA-09 to medication abortions was not reasonable because it produced no public health benefit, and indeed, was detrimental to achieving even its ostensible goals. And against this total lack of a benefit, there is substantial evidence that applying GA-09 to prohibit medication abortions posed a significant burden on the constitutional right to abortion, including by increasing the health risks, financial costs, and emotional toll associated with obtaining an abortion. It was thus entirely reasonable for the district court to conclude that the benefits of applying GA- 09 to medication abortions are “beyond all question” outweighed by its burden on the constitutional right to an abortion, Abbott II, 2020 WL 1685929, at *7, *9, thereby creating an unconstitutional undue burden under Casey. The majority faults the district court for not citing to the Petitioners’ exhibits purportedly containing contrary evidence on some of these points, calling this a failure to “carefully parse the evidence.” Majority at 2630. But “[i]t is the province of the district court to weigh conflicting evidence,” including by choosing which evidence to credit and which evidence to discount. R. S. by & through Ruth B. v. Highland Park Indep. Sch. Dist., 951 F.3d 319, 337 (5th Cir. 2020). It is not our role to second guess what would appear to be a reasonable evaluation of the evidence under commonplace circumstances, let alone on mandamus review. There is thus no basis to conclude that the district court’s reasoning produced “patently erroneous results” as to medication abortions. In re Lloyd’s Register N. Am., Inc., 780 F.3d at 290 (quoting In re Volkswagen of Am., Inc., 545 F.3d at 312). 2. The majority similarly concludes that the district court’s reasoning led to a patently erroneous result with regard to the TRO’s blocking application of GA-09 to those abortions in the 18-week category—that is, those abortions for women who would exceed the maximum gestational age to legally have an abortion other than in an ambulatory surgical center by the expiration of the executive order and who would, in their physicians’ judgment, be unable to obtain an abortion at one of these centers. Majority at 32-34. The majority first criticizes the district court for categorizing the application of GA-09 to women in this 18-week category as an “absolute ban on abortion” because, it contends, women falling into this category can, theoretically, still legally obtain an abortion. Majority at 32 (quoting Abbott III, 2020 WL 1815587, at *6). The majority’s argument is based on a theoretical legal possibility that is a practical impossibility. Many women in this category will not be able to obtain an abortion for a number of reasons. In reality, there are no ambulatory surgical centers that provide abortion care outside of Texas’s four largest metropolitan areas, see Hellerstedt, 136 S. Ct. at 2316, and thus many women in rural areas of Texas would need to secure transportation over a great distance and lodging in a metropolitan area in order to undergo the two-day procedure necessary for an abortion after the 18- week mark. See App. at 130-31. For many this will not be possible due to time constraints, financial limitations, health reasons, or any number of other factors. And there is evidence that, because of the buildup of need for abortion care during the time GA-09 is in effect, the delays associated with the resulting backlog may prevent many women who will be past the 18-week mark upon the expiration of the executive order from obtaining an abortion before the 22- week legal cutoff.[43] See, e.g., App. at 95. For these women, GA-09 is for all intents and purposes an absolute ban on abortion. And the majority offers no authority for the prospect that a law that theoretically leaves a legal path to abortion cannot as a practical matter function as the sort of absolute ban that violates “the essential holding of Roe v. Wade,” Casey, 505 U.S. at 846. In light of the balancing test for identifying an undue burden set forth in Casey, 505 U.S. at 874, and Hellerstedt, 136 S. Ct. at 2310, the outright ban on previability abortion might be thought of as the ultimate burden on the constitutional right to abortion which no benefit or interest yet identified can outweigh. In that C8.SG, 8. law that nominally allows abortion but places it functionally out of reach for a class of women is only slightly less of a burden, and it stands to reason that a truly compelling benefit would be required to justify it. We need not speculate what that benefit might be, though, because on the evidence in the record, it was reasonable for the district court to conclude that applying GA-09 to prohibit this class of abortions offered no benefit at all. Much of the evidence already recounted concerned not only the relative consumption of PPE and hospital capacity between a medication abortion and a continued pregnancy, but rather any pre-viability abortion and a continued pregnancy. See, e.g., App. at 135 (“By comparison, even if a provider of prenatal care reduces the scheduling of such care during the COVID-19 outbreak, it will still involve use of masks, sterile gloves, and potentially other PPE during multiple visits. A patient continuing pregnancy will thus require significantly more PPE than a patient presenting for abortion.”); App. at 375 (stating that “most prenatal and postpartum care is continuing” during the COVID-19 pandemic and cannot be done remotely before concluding that “requiring people to continue unwanted pregnancies utilizes more PPE and more hospital resources than abortion care”). It is true that, as the majority notes, the Supreme Court acknowledged that “increased driving distances do not always constitute an ‘undue burden.’” Hellerstedt, 136 S. Ct. at 2313. But the district court’s conclusion was not based on increased driving distance alone. The distance was “one additional burden, which, when taken together” with the forced postponement of abortion care, which created a significant backlog of need with a “virtual absence of any health benefit,” id., “beyond question” constituted an undue burden, Abbott II, 2020 WL 1685929, at *11. This conclusion was a reasonable interpretation of the evidence, and it therefore did not produce palpably erroneous results such that mandamus is appropriate.[44] CONCLUSION The present case is an excellent demonstration of the dangers of using the extraordinary remedy of mandamus to overmanage matters that are properly left to a district court’s discretion. In part because of the decisions of this court, the legality of abortion in Texas has changed no less than six times since the beginning of the COVID-19 pandemic. This court has expended substantial time and judicial resources in an effort to prevent interference with the state’s pandemic response at a most urgent time, only to instead contribute to a confusion that is likely more disruptive than the alleged harm it sought to prevent. Even today’s order will have little practical effect other than to briefly change the legality once more. Under GA-15, which takes effect at midnight on April 22, abortion legality in Texas will apparently change for an eighth time, as the Respondents have represented that all of their abortion care will fall into the new exception that exempts services provided by heath facilities that certify they will not draw upon any public supply of PPE. The majority again concludes that mandamus is appropriate to correct what it perceives as rampant abuses of discretion by the district court that produced patently erroneous results. As I have said, I strongly disagree with the majority’s critique of the district court’s work, and I do not believe that this case warrants mandamus relief. I therefore once again respectfully but emphatically dissent.
