Before JONES, SMITH, and HAYNES, Circuit Judges. JERRY E. SMITH, Circuit Judge: Attorney Andrew Willey wants to solicit legal work from already-represented criminal defendants in Harris County. But he fears that would violate a Texas anti-barratry law, so he sued for a preliminary injunction prohibiting the Harris County District Attorney from enforcing that law against him. The district court denied an injunction. Willey appeals, urging that prosecuting him for soliciting work from represented defendants would violate his First Amendment rights to free speech and association. We affirm because Willey has not shown that his claim is likely to succeed. I. A. Willey is motivated by his belief that appointed criminal defense attorneys in Harris County are pervasively inadequate, largely because they are overburdened. Willey planned to help by representing affected indigent defendants pro bono, but solely to challenge their existing attorneys’ adequacy. He initially targeted the clients of a frequently-appointed criminal defense attorney in Harris County (“Doe”) because Willey believed Doe was especially overburdened. Willey created two forms to distribute to Doe’s clients. The first was titled “Representation Affidavit.” Spanning eight pages and seventy-four questions, the document solicited information about Doe’s performance. It provided for affiants to express their “wish” that Doe be replaced and to declare, “I cannot afford to hire a different attorney and am stuck with [Doe] merely because I am too poor to afford anyone else.” The second form was titled “Limited Scope Of Representation Agreement.” It explained that Willey would not become “undersigned Defendant’s attorney of record” or “replace the court-appointed attorney.” Instead, Willey’s representation would occur only on any “motion for new court-appointed counsel, motion of rehearing as such, and writ(s) of mandamus to enforce such motions.” And Willey would not be compensated. Willey enlisted “volunteer investigators” to identify Doe’s appointed clients. With the volunteers’ help, he distributed his forms to 22 of Doe’s clients. That prompted the families of two of those clients to contact Doe to ask whether he was still their relatives’ lawyer. Those inquiries alerted Doe to Willey’s activities. Doe notified Harris County District Judge Amy Martin, who was presiding over criminal cases brought against some of the contacted defendants. After two attempts to persuade Willey to stop contacting Doe’s appointed clients, Judge Martin convened an “emergency hearing.” There, she recounted an earlier conversation with Kermit Johnson, one of the defendants whom Willey had contacted. Judge Martin explained, and Johnson agreed, that Johnson had asked to speak with the court because Willey had visited him in jail. Johnson further agreed that, during Willey’s visit, Johnson wasn’t “feeling really good,” had “memory issues,” and had taken psychotropic medication. Johnson did not wish to speak with Willey, but the jail staff forced him to attend the meeting. Then, Willey induced him to sign “some documents,” presumably the forms described above. Johnson later asked to see those documents, but Willey refused. Judge Martin explained that Johnson “has a fairly serious illness” and said, “If whoever interviewed him could not figure out that he was under the influence of psychotropic meds, that person has no business representing indigent criminal defendants.” She told Willey it was “a very, very bad idea to continue to try and solicit business, whether paid or unpaid[,] from defendants he knows are represented,” particularly by court-appointed attorneys. She promised Johnson that Willey would not contact him again. She warned that if Willey continued his behavior, she would “not be nearly so nice.” Willey eventually promised to stop contacting Doe’s clients. He interpreted Judge Martin’s comments as a threat that he would face criminal sanctions if he continued soliciting legal work from represented defendants. So, he put his plans “on hold.” B. Texas forbids some forms of what it calls “Barratry and Solicitation of Professional Employment.”[1] TEX. PENAL CODE § 38.12 (West 2013). Attorneys may not contact represented parties “with the intent to obtain professional employment” relating to legal representation in “a specific matter.” Id. § 38.12(d).[2] An attorney may violate Section 38.12(d) even if he seeks no economic benefit. See id. So Willey fears that even solicitations directed at limited, pro bono representation are forbidden. Willey brought this pre-enforcement suit[3] against the D.A. and requested preliminary and permanent injunctions forbidding his prosecution under Section 38.12. After a hearing, the district court denied Willey’s motion for a preliminary injunction.[4] Willey appeals that denial. Willey maintains that he is entitled to a preliminary injunction because his desired conduct is protected by the First Amendment. Since his exercise of constitutional rights is restrained, he says he is suffering an irreparable injury, and the balance of equities and the public interest favor injunction. He urges that he is likely to succeed on the merits because Texas has criminalized protected political speech and association, warranting strict constitutional scrutiny. Willey reasons that the law cannot survive strict scrutiny because it is under-inclusive in protecting any compelling state interest. II. We review the denial of a preliminary injunction de novo because determining whether “free speech rights have been infringed presents a mixed question of law and fact.” Byrum v. Landreth, 566 F.3d 442, 445 (5th Cir. 2009). We apply the familiar standard for assessing the propriety of a preliminary injunction: Willey must show “(1) a substantial likelihood of success on the merits; (2) a substantial threat of irreparable harm if the injunction is not granted; (3) that the threatened injury outweighs any harm that the injunction might cause to the defendant; and (4) that the injunction will not disserve the public interest.” Opulent Life Church v. City of Holly Springs, 697 F.3d 279, 288 (5th Cir. 2012). We do not reach the last three factors because Willey has not shown substantial likelihood that “the challenged law is incompatible with the First Amendment.” Texans for Free Enter. v. Tex. Ethics Comm’n, 732 F.3d 535, 538 (5th Cir. 2013). III. A. Willey says he wishes to offer to represent indigent defendants pro bono for religious and political reasons.[5] He thus claims to be “employing constitutionally privileged means of expression to secure constitutionally guaranteed civil rights” rather than “procur[ing] remunerative employment.”[6] For this appeal, we assume, as do both parties, that application of the anti-barratry law to his conduct must withstand strict constitutional scrutiny. See In re Primus, 436 U.S. 412, 432 (1978).[7] Under that standard, the D.A. must prove that the “restriction furthers a compelling interest and is narrowly tailored to achieve that interest.”[8] The Supreme Court has twice applied strict scrutiny to state attempts to restrict non-commercial attorney solicitation. First, in Button, 371 U.S. at 434, the Court held it unconstitutional for Virginia to prohibit NAACP lawyers from advising potential litigants of their rights and referring them “to a particular attorney.” Virginia failed to demonstrate a compelling interest “in the form of substantive evils flowing from petitioner’s activities.” Id. at 444. Second, in Primus, 436 U.S. at 414−21, the Court held it unconstitutional for South Carolina to discipline an attorney for advising a potential litigant that the ACLU would represent her pro bono. The state pointed to the “substantive evils of undue influence, overreaching, misrepresentation, invasion of privacy, conflict of interest, and lay interference that potentially are present in solicitation of prospective clients by lawyers.” Id. at 426. The Court did not deny that preventing those harms was a compelling interest but explained that the state had to prove that “any of [those] substantive evils . . . were present in this case.” Id. at 433. The Court was careful not to “foreclose carefully tailored regulation that does not abridge unnecessarily” freedom of speech or association. Id. at 439. Button and Primus set the table, but neither controls the outcome. The subjects of Willey’s solicitations were already represented. That distinction implicates interests different from the anti-solicitation rationales advanced previously.[9] Whether the First Amendment permits a state to criminalize Willey’s desired conduct is a question of first impression. Button and Primus help answer that question by establishing three principles: (1) the work Willey allegedly wishes to do is constitutionally protected speech and association; (2) restrictions on that conduct are strictly scrutinized; but (3) restrictions are permissible where carefully tailored to prevent substantive evils that a state proves are present in a particular case. Restrictions on speech rarely withstand strict scrutiny, but strict scrutiny is not “strict in theory, fatal in fact.”[10] B. The D.A. advances three compelling interests to justify the restriction: (1) “preserving the attorney-client relationship from damaging interference,” (2) avoiding “unnecessary delays and confusion in court proceedings” resulting from the solicitations, and (3) “reasonable regulation of the legal profession.” We address only the state’s interest in preserving attorney-client relationships because it sustains application of the anti-barratry statute to Willey’s conduct at this preliminary stage. 1. The D.A. fears that Willey’s solicitations, combined with his questionnaire and representation agreement, will damage indigent defendants’ relationships with their court-appointed counsel. She maintains that Willey confused at least two defendants about who was representing them. We share that concern and conclude that preventing such confusion is a compelling state interest. Compelling interests rarely “reduce to precise definition.” See Williams-Yulee, 575 U.S. at 447.[11] The Supreme Court has sometimes described compelling interests with synonyms such as “interests of the highest order”[12] and “vital state interest[s].”[13] But it has offered no general theory of what makes a state interest “compelling.” We are left to infer the general from the particular. Interests previously held compelling enough to satisfy strict scrutiny include “safeguarding ‘public confidence in the fairness and integrity of the nation’s elected judges,’”[14] combating terrorism,[15] protecting election integrity,[16] remedying the effects of past unconstitutional race discrimination,[17] and attaining “a diverse student body” in post-secondary education.[18] The question is whether preventing interference with indigent defendants’ understanding of their representation belongs on that list. Texas’s proffered interest implicates its constitutional obligations. The state must provide indigent defendants with “reasonably effective assistance” of counsel. Washington, 466 U.S. at 687. Solicitation of a client by another attorney threatens an appointed lawyer’s ability to provide reasonably effective assistance because it can confuse a defendant about who represents him. A defendant who thinks someone else represents him may not provide his attorney of record with critical information about the case. That can lead to that attorney’s failure to pursue evidence or timely raise a key defense. Both can constitute ineffective assistance.[19] Of course, Texas is not obligated to criminalize attorney solicitations of represented defendants.[20] But that observation does not diminish the importance of the interest Texas serves. States have some latitude to determine the best way to structure the legal process while meeting their constitutional obligations. For instance, in Williams-Yulee, 575 U.S. at 448, Florida could have avoided the harms it sought to address with a speech restriction by switching from a partially elected judiciary to an appointed one. But the Court did not compel that choice and instead explained that a “[s]tate’s decision to elect its judges does not require it to tolerate” risks that the public will lose confidence in the integrity of its judiciary. Id. So too may Texas decide to focus its efforts to provide effective assistance to indigent defendants on establishing a single, unmolested attorney-client relationship. A state may choose to ward off constitutional injuries at the source instead of waiting for them to materialize. Still, a constitutional obligation is not ipso facto a compelling interest.[21] When relying on constitutional compliance as the source of a compelling interest, the Supreme Court has not stopped at identifying the obligation. Instead, it has balanced the relative importance of the competing interests. For instance, in one line of cases, it held that compliance with the Establishment Clause can justify content-based restrictions on speech, but it cautioned that the same interest may not permit viewpoint discrimination. See Good News Club v. Milford Cent. Sch., 533 U.S. 98, 112–13 (2001). That distinction can be explained only by the inference that the Court balanced harms and ranked the Establishment Clause more important than the harm caused by content discrimination but less important than that caused by viewpoint discrimination. The need for balancing further follows from the asserted conflict among constitutional values. It would be arbitrary to say that any constitutional obligation is compelling enough to sustain infringement of any constitutional right. If that were true, the outcome of strict-scrutiny analysis could turn solely on which constitutional value a state initially chooses to give preference to. If two states make opposing choices in the same context, and both are challenged, the cases could come out differently despite implicating the same constitutional conflict. To avoid such arbitrariness, if there is a true conflict between two constitutional values such that one must give way to the other, their relative importance must be assessed.[22] That is why the Supreme Court has sometimes described strict scrutiny as a “balancing test” and a compelling interest as a “subordinating” one, which implies comparative analysis. See Emp. Div., Dept. of Hum. Res. v. Smith, 494 U.S. 872, 882–83 (1990); NAACP v. Alabama, 357 U.S. 449, 463 (1958) (citation omitted). Thus, we must determine whether the state’s asserted interest in ensuring indigent defendants reasonably effective assistance of counsel by protecting attorney-client relationships can subordinate the rights to free speech and free association. We conclude that it can. The most analogous state interest that has been evaluated by the Supreme Court is that in promoting public confidence in the integrity of the judiciary. See Williams-Yulee, 575 U.S. at 444–48. A state’s compelling interest in the “integrity of its election process” is also relevant. See Burson, 504 U.S. at 199 (quotation omitted). Those interests share a concern with the fairness and legitimacy of government. A state has no higher interest.[23] Little else matters if a state cannot persuade its citizens that it wields legitimate authority. See, e.g., Luther v. Borden, 48 U.S. (1 How.) 1, 34–38 (1849) (recounting the chaos resulting from a dispute about the legitimacy of Rhode Island’s post-colonial government). That is why states have been permitted to limit expression to the extent necessary to ensure that their judiciaries are impartial and their elections are free and fair. The right to effective assistance of counsel has comparable gravity. Such assistance is “fundamental and essential to fair trials” and necessary to produce “just results” in our adversarial system. Gideon v. Wainwright, 372 U.S. 335, 344 (1963); Washington, 466 U.S. at 685. If attorney solicitation impedes effective assistance as previously discussed, a case may be adjudicated following an incomplete consideration of the merits of the charge. Even if that error is ultimately corrected on review, courts’ ability to do justice is undermined in public perception and reality. So Texas’s interest is ultimately about the fairness of its criminal adjudications. That interest ranks alongside the others concerning a state’s ability to ensure the legitimacy of its coercive authority. Since those interests are of a piece, we need not independently balance Texas’s asserted interest against Willey’s First Amendment rights. The Supreme Court has already determined that such interests can subordinate freedom of speech if necessary. See Williams-Yulee, 575 U.S. at 448; Burson, 198–99. It follows that the asserted state interest here is similarly subordinating. It may sustain application of the anti-barratry law to Willey’s conduct if the law is narrowly tailored to its end. 2. A speech restriction is narrowly tailored to a compelling state interest where it does no more “than is necessary to further” the interest. Knowles v. City of Waco, 462 F.3d 430, 435 (5th Cir. 2006) (quoting Ward v. Rock Against Racism, 491 U.S. 781, 801 (1989)). The “essence of narrow tailoring” is to “focus[ ] on the source of the evils the [government] seeks to eliminate . . . without at the same time banning or significantly restricting a substantial quantity of speech that does not create the same evils.” Ward, 491 U.S. at 799 n.7. Here, the anti-barratry law must not go further than necessary to prevent indigent criminal defendants from being confused about who represents them. A narrowly tailored restriction must also advance the asserted interest well enough to prove that the interest is genuine and not a post hoc rationalization. Defects of that type are termed “under-inclusivity.” The Supreme Court has often found under-inclusivity antithetical to narrow tailoring.[24] Nonetheless, under-inclusivity is not per se fatal under strict scrutiny. Williams-Yulee, 575 U.S. at 448–50. After all, it is odd to suggest that a First Amendment defect might be cured by restricting more speech. Id. at 448. So it is not enough to point out that a state did not “address all aspects of a problem in one fell swoop.” Id. at 449. Instead, under-inclusivity tests the sincerity of the state’s interest. We ask whether the restriction is “riddled with exceptions” or “so woefully underinclusive as to render belief in [its stated] purpose a challenge to the credulous.” Id.; White, 536 U.S. at 780. Under-inclusivity is the principal ground on which Willey attacks the anti-barratry statute. He finds it under-inclusive in two ways. First, he says the portion of the statute that could be applied to his conduct can proscribe only true speech—that is, solicitation of represented clients based on the inadequacy of their appointed attorney when that attorney truly is inadequate. Thus, Willey reasons that the statute serves no legitimate state interest. Second, Willey posits that attorney-client relationships can be harmed in myriad ways that aren’t covered by the statute. Attorneys can say anything they like to damage the same relationships so long as they do not “seek to become [a defendant's] lawyer in the matter in which they are represented.” Addressing Willey’s first argument requires statutory explication. Section 38.12(d) prohibits six types of actions by attorneys and other licensed professionals. On top of the prohibition on solicitation for legal services that is directly relevant here,[25] it forbids communications that “involve[ ] coercion, duress, fraud, overreaching, harassment, intimidation, or undue influence”[26] or “contain[ ] a false, fraudulent, misleading, deceptive, or unfair statement or claim.”[27] Like the anti-solicitation provision, these prohibitions apply to any action intended to “obtain professional employment,” even if not for profit. Id. § 38.12(d)(2). Willey thus tries to characterize the anti-solicitation provision as sur-plusage vis-à-vis the state’s interest in protecting attorney-client relationships. He divides attorneys’ hypothetical statements to represented parties into two categories: “potentially false statements” and “potentially true statements.” He reasons that any “potentially false statement” is already covered by Sections (d)(2)(E) and (F) because it would be, for instance, misleading, fraudulent, or overreaching. So he concludes that the only independent force of the anti-solicitation provision is to prohibit “potentially true statements.” He says a state has no legitimate interest in foreclosing attempts to terminate deficient representation. That dichotomy is way off the mark. For one thing, “potentially false” and “potentially true” mean more or less the same thing in this context. Willey’s equivocation about the truth of these hypothetical statements arises from the indeterminacy of whether representation is adequate. In Washington, 466 U.S. at 687−89, the Court adopted a consciously vague “reasonableness” standard, stating that “[m]ore specific guidelines [were] not appropriate” because the inquiry requires a consideration of “all the circumstances.” Whenever all circumstances must be considered, reasonable minds will differ. One lawyer may genuinely believe that another has rendered inadequate assistance and say so, honestly and without exerting undue influence or overreaching, even where courts and other lawyers would disagree—especially if that lawyer is predisposed to see the specter of inadequacy. For that reason, anything concerning a statement’s truth or falsity is a poor lens through which to view the state’s interest. Willey neglects to consider that Texas is legitimately interested in forbidding attorney solicitation of represented parties even where lawyers do not consciously coerce or mislead. If that is true, the anti-solicitation provision is not duplicative, as Willey claims. That leads to his second argument. Willey is right that attorneys can say anything they like to represented clients without violating the anti-solicitation provision so long as they do not “seek to become their lawyer in the matter in which they are represented.” That could mean an attorney is free to opine that a represented party’s lawyer is constitutionally inadequate, provided that attorney does not run afoul of the anti-fraud and anti-coercion provisions. But characterizing that as under-inclusivity misapprehends the nature of the state’s interest. The state’s interest is in preventing confusion that damages relationships between appointed counsel and indigent defendants. The potential for confusion is uniquely present where an attorney approaches a represented defendant and offers to become his lawyer, even for a limited purpose. If, as Willey suggests, an attorney merely “approach[es] represented defendants and explain[s] to them the deficiencies of their current lawyers,” those defendants are likely to raise those concerns with their current lawyers or a court. That would do little or no harm. But if Willey induces them to sign forms filled with legalese, including one titled “Limited Scope of Representation Agreement,” they likely will think he represents them generally and may withhold vital information from their counsel of record, leading to the harms already described. No matter how carefully drafted, Willey’s language limiting the scope of his representation is not likely to be communicated adequately to each of the recipients of his solicitations. Those untrained as lawyers may fail to appreciate the difference between “writ(s) of mandamus to enforce such motions” and other forms of representation. All they are likely to hear is that their old lawyer was no good, so they have a new one. That is not conjecture. As Primus requires, the state has demonstrated, as far as possible at this preliminary stage, that Willey’s conduct caused that precise type of confusion.[28] Family members of two of the twenty-two defendants whom Willey contacted were confused about whether Doe still represented them. Willey provided his forms to Kermit Johnson against his will, in jail, late at night, while he was struggling with mental illness and under the influence of psychotropic medication. It is hard to think of a better way to confuse a defendant about who represents him. Willey would have us draw a different conclusion from these facts. At oral argument, his counsel suggested that at least one criminal defendant was confused “because he never hears from his lawyer,” not because of Willey’s actions.[29] In other words, he says any confusion was Doe’s fault, not Willey’s. That is not supported by the record: It would not explain why everyone involved immediately connected the defendants’ confusion with Willey or why the families contacted Doe only after Willey’s solicitations.[30] And it entirely ignores Willey’s conduct involving Kermit Johnson, whose confusion couldn’t possibly be attributed to Doe. But likelihood is all we can judge at this preliminary stage and on this record. It will be for the district court to find facts in the first instance and for the D.A. ultimately to prove that Willey’s actions caused confusion. If that view of the facts is borne out, prohibiting Willey’s conduct precisely addresses the state’s compelling interest. There is a special harm created where an attorney solicits work relating to a particular legal matter and the subject of the solicitation already has an attorney—all the more so if the subject is especially vulnerable to confusion. That harm is not created by merely expressing an opinion about a defendant’s current lawyer. Hence, the anti-barratry law Willey challenges is the “essence of narrow tailoring” as it relates to the state’s compelling interest in protecting relationships between indigent defendants and appointed counsel. Ward, 491 U.S. at 799 n.7. It prevents exactly the targeted harm with no exceptions or carve-outs. In sum, because the anti-barratry law is likely narrowly tailored to a compelling government interest implicated by Willey’s conduct, Willey has not shown that he is likely to succeed on the merits of his First Amendment claim. We therefore decline preliminarily to enjoin the D.A. from enforcing the statute against him. IV. Finally, Willey asks us to assign this case to a different district judge on remand. During a hearing on the request for a preliminary injunction, the district judge erroneously referred to Willey’s organization Restoring Justice as a “defunct charity” and asked several questions about the organization and its tax status, even though it was not a party. Evidently, the judge confused Restoring Justice with another organization called “Restore Justice.” Willey calls that an “improper inquiry” and a “fixation on the tax compliance of the wrong non-profit,” which the judge never indicated would be set aside. Judges must be disqualified “in any proceeding in which [their] impartiality might reasonably be questioned.” 28 U.S.C. § 455(a). And appellate courts may reassign cases to a different judge on remand under their authority to “require such further proceedings to be had as may be just under the circumstances.” 28 U.S.C. § 2106; Liteky v. United States, 510 U.S. 540, 554 (1994). But reassignment is an “extraordinary” and “rarely invoked” power. In re DaimlerChrysler Corp., 294 F.3d 697, 700 (5th Cir. 2002) (quoting Johnson v. Sawyer, 120 F.3d 1307, 1333 (5th Cir. 1997)). We have used two independent tests to decide whether to reassign. The first test considers three factors: whether the original judge would reasonably be expected upon remand to have substantial difficulty in putting out of his or her mind previously-expressed views or findings determined to be erroneous or based on evidence that must be rejected, (2) whether reassignment is advisable to preserve the appearance of justice, and (3) whether reassignment would entail waste and duplication out of proportion to any gain in preserving the appearance of fairness. Id. at 700–01. The second test is whether the facts “might reasonably cause an objective observer to question the judge’s impartiality.”[31] The tests are “redundan[t]” because “ the second factor of the first test is virtually identical to the single question the simpler test asks,” and “[t]he result of each test has been always the same.” United States v. Khan, 997 F.3d 242, 249 n.4 (5th Cir. 2021). This case does not merit reassignment under either test. The district court’s mistake does not manifest bias to an objective observer because there were legitimate reasons to ask about Restoring Justice, which is mentioned six times in Willey’s complaint. And there was discussion in the record about whether Willey might evade his agreement with Doe by resuming his activities under the name of Restoring Justice. It was justifiable to ask about which parties were properly before the court, the relationships between them, and their legal capacities to act. Willey further reasons that the district judge must have been intentionally “searching for potentially negative information” to unearth anything about Restore Justice’s tax compliance. That is unduly speculative. We have no information in the record about how the judge found the information. We will not assume the worst, nor would an objective observer.[32] Finally, there is no reason to think the district judge will “have substantial difficulty in putting [this] out of his . . . mind.” DaimlerChrysler, 294 F.3d at 700. It is now clear that Restoring Justice is not defunct, facing tax difficulties, or a party to this case. We expect this issue to have no bearing whatsoever on further proceedings. Accordingly, the orders denying the preliminary injunction and the recusal of the district judge are AFFIRMED. The case remains in the district court for further proceedings, which may now resume. EDITH H. JONES, Circuit Judge, concurring: In this as-applied challenge, we are called upon to determine only whether Texas’s criminal anti-barratry statute may constitutionally be applied to Willey’s conduct as described. I think it can, and therefore concur in the judgment and in the majority’s narrow tailoring analysis. With respect to the state’s compelling interest, the state and the public have a grave interest and constitutional duty to provide free defense counsel to indigent criminal defendants. Gideon v. Wainwright, 372 U.S. 335, 339-340, 83 S. Ct. 792, 794 (1963) (citations omitted).[33] On the facts here alleged, it appears that Willey undertook deliberately to interfere with a particular court-appointed attorney’s representation of multiple clients. It is also clear that several clients or their families were unnerved and confused by his intrusive inquiries about the appointed attorney’s representation. The state court properly indicated concern as to whether, inter alia, Willey had caused the defendants to reveal attorney-client confidences or, contrarily, had caused the defendants to resist full cooperation with appointed counsel. Either way, and at least for these reasons, the state has a compelling interest in sanctioning Willey’s “speech” that not only had a tendency but, as pled, actually did interfere with the attorney-client relationship provided to these defendants. As this record stands, Willey has not shown a likelihood of success on the merits.
