Before SMITH, WIENER, and SOUTHWICK, Circuit Judges. JERRY E. SMITH, Circuit Judge: GenMa is a power company that, long ago, leased two coal-fired power plants from the Lessors. To comply with those leases, GenMa paid NFC $130 million to insure the Lessors up to that sum if GenMa didn’t pay rent. Too late, NFC realized it had promised the Lessors more than $130 million. The Lessors forced NFC to honor its promise, and NFC sued GenMa and others for its losses. GenMa removed NFC’s claims to the Southern District of New York, which then transferred those claims to a bankruptcy court in Texas. After losing there and at the district court, NFC appeals. It says that its claims against GenMa should return to New York state court because the federal court lacked jurisdiction or because federal law required abstention. NFC also insists, pressing four contract-law theories, that GenMa must cover NFC’s losses. We disagree. The district court had jurisdiction; abstention was not required; and NFC’s claims lack merit. I. The defendant-appellee is GenOn Mid-Atlantic—GenMa for short. GenMa operates several power plants in Maryland.[1] The firm is a subsidiary of GenOn Energy and NRG Energy, one of the largest retail power companies in America. About two decades ago, GenMa leased two coal-fired power plants from various entities, whom we will call the Lessors.[2] In those leases, GenMa made two promises. First, GenMa agreed not to grant any liens on its assets. Second, GenMa agreed to obtain credit for the Lessors to secure six months’ worth of rent payments, but GenMa would not grant liens on its own assets to collateralize that credit. That restriction ensured that the Lessors’ drawing that credit would not diminish GenMa’s ability to pay rent. For years, GenMa got that credit support from its corporate parent, GenOn Energy (“GenOn”). Things changed in 2016. GenMa had obtained, for the Lessors, letters of credit from JP Morgan Chase. But those letters were expiring soon, and GenOn, which had backed them, was struggling to pay its own bills. Because of GenOn’s distress, NRG, which owned GenOn and GenMa, declined to back new letters of credit. And GenMa had no other investment-grade affiliates that could guarantee its rent. Needing another way to post credit for the Lessors, GenMa turned to its strengths. At the time, it was financially stable and had some cash on hand, so it decided to buy a letter of credit. GenMa took its proposal to Natixis Funding Corporation (“NFC”), the plaintiff-appellant. A subsidiary of a large French bank, NFC sells letters of credit and other structured financial products. Mindful of the leases, GenMa insisted on “structur[ing]” its purchase “ as a prepayment” and not as “a cash collateralized instrument.” GenMa even shared its leases with NFC’s team, highlighting the “qualifying credit support requirements.” Both sides engaged sophisticated representatives for the negotiations. Two months later, after exchanging multiple drafts, GenMa and NFC inked the Payment Agreement, under which GenMa paid NFC $130 million plus a $1.4 million letter-of-credit fee. The $130 million sum reflected the greatest amount of credit that GenMa had to provide the Lessors in one lease period. In exchange, NFC promised to obtain letters of credit from its New York affiliate, which we will call Natixis, for the Lessors. If the letters went undrawn, NFC would pay up to $130 million in rent to the Lessors on GenMa’s behalf. If the letters were drawn, NFC would reimburse Natixis for those draws. Reflecting GenMa’s need to avoid creating a lien, which its leases proscribed, the Payment Agreement repeatedly disclaimed GenMa’s interest in the $130 million payment to NFC. That payment, the Agreement stressed, was “in full,” upfront, and “irrevocabl[e].” GenMa, it continued, renounced any “interest, claim, reversionary or residual interest” in the payment. The Agreement also assured that NFC would bear all risk and reward on the payment. NFC would receive “any returns, interest, gains[,] or other earnings” that accrued and would assume all risk of the payment’s loss. And the parties “understood and agreed that . . . the [$130 million] has been indefeasibly paid by [GenMa] to NFC.” (Emphasis added.) The Agreement protected NFC in three relevant ways. First, the Agreement capped NFC’s duty to pay rent or to allow credit draws at each Lessor’s share of $130 million—GenMa’s payment amount. The Agreement called that share the “Excess Capacity of Lessor.” Letter-of-credit draws by, or lease payments to, a Lessor would reduce its “Excess Capacity.” When that capacity reached zero, NFC’s duties to pay rent or to provide letters of credit to that Lessor would cease. Second, GenMa warranted that the Payment Agreement didn’t breach its promises to the Lessors, including the promise not to incur liens to secure their credit support. If GenMa breached that warranty, it would indemnify NFC against costs incurred to enforce its rights under the agreement. Third, GenMa agreed to indemnify NFC against losses from “judicial proceeding[s] . . . brought or threatened” by third parties. But that indemnity excluded, among other things, “any reimbursement obligation to . . . any . . . Person in respect of any” lease payment or letter-of-credit disbursement. After NFC and GenMa executed the Payment Agreement, Natixis issued letters of credit to the Lessors. But those letters covered all lease periods—over $2 billion in rent—and did not cap draws as the Agreement allowed. Immediately, several Lessors questioned whether the Natixis letters complied with the leases’ credit-support requirement. They objected that GenMa had used its own cash to buy the letters. In early 2017, five Lessors noticed GenMa’s default and directed their representative, the indenture trustee, to draw $125 million on the JP Morgan letters of credit before they expired. The trustee duly drew the funds and escrowed the proceeds. After those draws, GenMa and the Lessors sued each other in New York state court, disputing whether the Natixis letters of credit complied with GenMa’s leases.[3] The Lessors’ suit included, as defendants, GenMa’s corporate parents, GenOn and NRG. In June 2017, days after the Lessors sued, GenOn and several subsidiaries—but not GenMa—filed for bankruptcy in the Southern District of Texas. GenOn soon moved the bankruptcy court to estimate the value of the Lessors’ claims against it at zero dollars. That motion required the bankruptcy court to examine the credit-support requirement in GenMa’s leases. After discovery, the bankruptcy judge held a ten-day trial and granted GenOn’s motion, valuing the Lessors’ claims against GenOn at zero dollars. In December, the parties to the state-court suits—notably GenMa, GenOn, and the Lessors—outlined the terms of a settlement. They agreed to release all claims against each other. In exchange, GenMa promised to pay off the Lessors’ debt on the leases with cash and debt, keeping at least $25 million in cash on hand.[4] The bankruptcy court then enshrined the settlement term sheet in GenOn’s reorganization plan. The settlement, the court explained, thwarted “complex and protracted litigation” that could “derail the Debtors’ reorganization efforts.” Flashback to June, just after GenOn’s bankruptcy filing. When GenMa’s rent came due at the end of that month, GenMa asked the Lessors’ trustee to use the proceeds of the JP Morgan letters, which remained escrowed, to cover its rent. The trustee refused and, when GenMa did not pay rent, drew $125 million on Natixis’s letters. Natixis honored those draws, which NFC duly repaid. At that point, NFC and Natixis realized that they were overexposed. They already had paid the Lessors $125 million. Yet they had promised the Lessors tens of millions more in credit support for the next lease period. So if the Lessors drew on the letters again, the liability of NFC and Natixis could far exceed $130 million. Alarmed, Natixis terminated the letters of credit, giving sixty days’ notice as the letters required. But before that time could run, the Lessors tried to draw another $50 million on the letters. Natixis refused to honor those draws. Instead, Natixis and NFC sued the Lessors, the indenture trustee, and GenMa, in New York state court. This appeal concerns only NFC’s claims against GenMa.[5] NFC asserted two contract claims: NFC first claimed that GenMa had breached its warranty that their Agreement did not break GenMa’s promises to the Lessors—namely, GenMa’s promise not to use its own assets to secure the Lessors’ credit support. NFC next asserted a breach of the implied covenant of good faith and fair dealing. Contending that those claims “related to” GenOn’s bankruptcy, 28 U.S.C. § 1334(b), GenMa removed them to the Southern District of New York under § 1452. GenMa then moved to transfer those claims to the Southern District of Texas for resolution before the bankruptcy judge managing GenOn’s reorganization. GenMa pointed out that the bankruptcy judge had evaluated the Lessors’ claims regarding the same lease provisions during the claims-estimation trial. “No court,” GenMa reasoned, “is more familiar with the legal and factual issues underlying the Removed Claims than the Texas Bankruptcy Court.” NFC moved to remand to state court. It pointed out that GenOn, not GenMa, was bankrupt, and that GenOn had confirmed its reorganization plan. NFC also claimed that the dispute between it and GenMa—a contract dispute between non-debtors—would not impact GenOn’s assets or reorganization. And even if there were jurisdiction, NFC argued, the district court could not exercise it. Section 1334(c)(2) requires a district court with related-to-bankruptcy jurisdiction to abstain from hearing claims where— among other requisites—”an action” regarding those claims “could not have been commenced” in federal court and where a state court could have “timely adjudicated” those claims. According to NFC, those requisites were met. GenMa contested both points. Against remand, GenMa observed that GenOn had not completed its confirmed reorganization plan and so had “not yet emerged from bankruptcy.” GenMa warned that NFC’s claims threatened to torpedo GenOn’s reorganization by “disrupt[ing] the careful balance struck” in the settlement between the Lessors, GenMa, and GenOn. Against abstention, GenMa stated that diversity jurisdiction would exist had the removed claims started in federal court and that the bankruptcy court, already familiar with the dispute, could resolve it faster than could the state court. At a hearing on those motions, the parties clashed over whether the removed claims could affect GenOn’s reorganization. Observing that the bankruptcy court would be “in the best position” to answer that question, the district court, Judge George Daniels, directed the parties to solicit the bankruptcy court’s views. The parties duly moved the bankruptcy judge for a hearing. There, Chief Judge David Jones opined that NFC’s claims threatened a “huge potential effect” on GenOn’s reorganization, that there is “a very close tie between the resolution of [NFC's] claims and the ultimate consummation of [GenOn's] plan,” and that his court had jurisdiction to resolve NFC’s claims. Persuaded, Judge Daniels denied NFC’s motion to remand, ruled abstention unnecessary, and transferred NFC’s claims to the Southern District of Texas. Over NFC’s objection, the Texas district court, Judge Keith Ellison, left pretrial matters to the bankruptcy judge. At the bankruptcy court, NFC twice amended its complaint. The operative one states four claims against GenMa: (1) contract reformation; (2) breach of the implied covenant of good faith and fair dealing; (3) breach of warranty; and (4) contractual indemnification. GenMa soon moved to dismiss that complaint for failure to state a claim. From the bench, Chief Judge Jones said that he would grant that motion as to the reformation and implied-covenant claims but would allow the plaintiffs’ other claims to proceed. After discovery, GenMa sought summary judgment on the surviving claims. NFC opposed. In a written report and recommendation, the bankruptcy judge advised the district court to give GenMa summary judgment on the breach-of-warranty and indemnification claims. Those claims failed, he explained, because (1) the Agreement did not breach the credit-support requirement of GenMa’s leases and (2) the indemnity provisions in that Agreement did not cover “litigation that NFC commenced and lost.” Chief Judge Jones again advised dismissing NFC’s reformation and implied-covenant claims. He attributed NFC’s losses to Natixis’s “careless[ ]” drafting of the letters of credit and observed that the Agreement imposed on NFC “ all risks [of ] a draw” under those letters. After the district court adopted that report and entered a take-nothing judgment, NFC appealed. It questions our jurisdiction and insists on abstention. Alternatively, NFC seeks to revive its claims against GenMa. We reject NFC’s contentions and affirm the district court’s judgment. To explain why, we turn first to jurisdiction, then to abstention, and finally to NFC’s contract claims. II. Jurisdiction comes first,[6] and we review it de novo.[7] GenMa removed this dispute to federal court, claiming it “related to” GenOn’s bankruptcy. 28 U.S.C. § 1334(b). We must decide whether that relationship existed, and we conclude that it did. But before we explain why, we must sketch the limits of our power to hear cases like this one. A. Federal district courts may hear “all civil proceedings . . . related to” bankruptcy cases. Ibid. A proceeding relates to a bankruptcy case if “the outcome of that proceeding could conceivably have any effect” on the debtor’s estate. Bass v. Denny (In re Bass), 171 F.3d 1016, 1022 (5th Cir. 1999) (quotation omitted). “Related-to jurisdiction” thus includes “any litigation” that “could alter the debtor’s rights, liabilities, options, or freedom of action or could influence the administration of the bankrupt estate.” Collins v. Sidharthan (In re KSRP, Ltd.), 809 F.3d 263, 266 (5th Cir. 2015) (cleaned up). That jurisdiction is broad, but it narrows once the debtor confirms its reorganization plan. Confirmation dissolves the debtor’s estate and, with it, bankruptcy jurisdiction, except “for matters pertaining to the implementation or execution of the plan.” Craig’s Stores of Tex., Inc. v. Bank of La. (In re Craig’s Stores of Tex., Inc.), 266 F.3d 388, 390 (5th Cir. 2001). Craig’s Stores first set forth that standard. Craig’s and the Bank of Louisiana had a financing arrangement that persisted through Craig’s bankruptcy. After reorganizing, Craig’s sued the Bank for damages that arose only after the date its plan was confirmed. The district court vacated for want of related-to jurisdiction. Affirming that judgment, we noted several key facts: No “claim” was “pending between the parties as of the date of the reorganization.” Id. at 391. “[N]o facts or law deriving from the reorganization or the plan was necessary” to Craig’s claim against the Bank. Ibid. And the case before us had “nothing to do with any obligation created” by Craig’s reorganization plan. Ibid. The panel also dismissed Craig’s insistence that the “status of its contract with the Bank will affect its distribution to creditors under the plan,” because “the same could be said of any other post-confirmation contractual relations in which Craig’s is engaged.” Ibid. We later distilled Craig’s Stores into three factors relevant to jurisdiction over post-confirmation disputes. In re Enron Corp. Sec., 535 F.3d 325, 335 (5th Cir. 2008). The first factor is whether the dispute “principally dealt with post-confirmation relations between the parties,” or instead arose from pre-confirmation conduct. Ibid (quotation omitted). The second is whether the claims were brought before confirmation. Ibid. The third is whether any “facts or law deriving from the reorganization or the plan were necessary to the claim.” Ibid. (quotation omitted and alteration adopted). Those factors are a useful heuristic, but only the Enron court has applied them. The rest of our decisions instead ask Craig’s overarching question: Does the dispute “pertain to the implementation or execution” of the debtor’s reorganization plan? U.S. Brass Corp. v. Travelers Ins. Grp. (In re U.S. Brass Corp.), 301 F.3d 296, 304 (5th Cir. 2002) (quoting Craig’s Stores, 266 F.3d at 391).[8] So it is to that question we turn. B. Our jurisdiction hinges on one question: Did NFC’s claims “pertain to the implementation or execution” of GenOn’s reorganization plan, Craig’s Stores, 266 F.3d at 391 (cleaned up), when GenMa removed NFC’s claims to federal court?[9] We agree with GenMa: The answer is yes. 1. This case is closer than our usual related-to fare. This is not a situation in which a bankruptcy court seeks to enforce its orders or to block alleged violations of a debtor’s bankruptcy-law rights. Galaz v. Katona (In re Galaz), 841 F.3d 316, 322–23 (5th Cir. 2016). Nor is this a dispute over the meaning of provisions of the reorganization plan. U.S. Brass, 301 F.3d at 305. In those cases, related-to jurisdiction is clear. This case instead lives at the limit of related-to jurisdiction. NFC and GenMa are non-debtors disputing state-law claims. As NFC observes, neither party “was a debtor or creditor in the GenOn bankruptcy, and none of the debtors were parties” to the Payment Agreement at the center of this case. But those facts do not foreclose related-to jurisdiction. This court has held that related-to jurisdiction may extend to non-debtors’ state-law disputes after the debtor confirms its plan. Feld v. Zale Corp. (In re Zale), 62 F.3d 746, 757–59 (5th Cir. 1995). In Zale, the debtor (Zale) confirmed a plan. Later, Zale, its primary director-and-officer insurer (Cigna), and others settled threatened suits against the debtor’s former directors. The settlement included a third-party release—a provision purporting to block anyone, including nonparties to the settlement, from bringing any claims against the settling parties. And Cigna agreed to pay Zale the full limit of its primary D&O insurance policy in exchange for Zale’s indemnifying it against any claims regarding the settlement. But that settlement rankled NUFIC, Zale’s excess D&O liability insurer, which had been excluded from the settlement yet anticipated liability under its excess policy. Wishing to sue Cigna on various tort and contract theories, NUFIC appealed the settlement’s injunctive provisions. This court framed the question as “whether the bankruptcy court had jurisdiction over an attempt to enjoin actions between . . . NUFIC and Cigna.” Zale, 62 F.3d at 755. The panel concluded that the bankruptcy court lacked jurisdiction over the tort claims against Cigna. All parties to the proposed claims were non-debtors, the panel explained. And if NUFIC prevailed, Cigna would pay any damages from its “other assets,” not from proceeds of the debtor’s primary D&O policy. Ibid. The panel dismissed Zale’s promised indemnity as a basis for jurisdiction because the tort claims concerned only Cigna’s acts, not Zale’s, and because NUFIC could not have brought the tort claims against Zale “even indirectly.” Id. at 756. But the panel held that there was jurisdiction over NUFIC’s contract claims, for two reasons: First, Zale’s reorganization plan assumed that some money from the Cigna policy would enter Zale’s estate, an assumption that the settlement confirmed. Id. at 758. And if successful, NUFIC’s claims— that Cigna “improperly bypassed the limits on its policy and . . . shifted liability to NUFIC’s excess policy,” id. at 757 n.29—could change how much Zale’s estate could recover from the Cigna policy. Second, “suits over the Cigna policy” would “tie up the policy assets and other assets” of the litigation trust created by the reorganization plan, inhibiting Zale’s reorganization. Id. at 759 (cleaned up). Zale shows that we have jurisdiction here. To see why, recall the “instrumental” settlement between GenOn, GenMa, and the Lessors. The parties to that settlement agreed to release all claims relating to the Natixis letters of credit. In exchange, GenMa pledged to pay off the Lessors’ debt on the leases with a combination of cash and debt. The cash would come from GenMa’s cash reserves, which, under the settlement, could not drop below $25 million after the payoff. After the parties forged that tentative framework, the bankruptcy court enshrined it in the plan, finding that the proposal, if effected, would thwart “complex and protracted” litigation that could “derail the Debtors’ reorganization efforts.” When GenMa removed NFC’s claims to the Southern District of New York, that settlement was not yet final, and NFC had by then asserted more than $34 million in damages against GenMa. That brings us to the link between Zale and this case: In Zale, the dispute between NUFIC and Cigna risked disrupting Zale’s reorganization by threatening Zale’s recovery from and access to the Cigna policy funds. Here, NFC’s claims risked the same disruptions: GenMa had pledged to pay the Lessors lots of money and to keep specified cash reserves as part of a global settlement between several parties to GenOn’s restructuring. By threatening GenMa’s ability to fulfill those commitments, NFC’s claims pertained to “the implementation and execution” of that crucial settlement, which was part of GenOn’s plan. Craig’s Stores, 266 F.3d at 390. So we have related-to jurisdiction. 28 U.S.C. § 1334(b). Though Zale preceded Craig’s, it has no less force here. This court has found post-confirmation jurisdiction over suits between non-debtors after Craig’s Stores. Take Biloxi Casino Belle,[10] for example. There we asked whether a title-insurance policy covered a security interest in a debtor’s casino boat. Id. at 497. That clash “pertain[ed] to the implementation or execution of the [bankruptcy] plan,” we explained, because one of the parties to that suit had assigned its recovery to a trust created by the debtor’s plan “for the benefit of unsecured creditors.” Id. at 496 n.4.[11] So both before and after Craig’s Stores, this court has accepted that related-to jurisdiction can exist over state-law disputes between non-debtors. Exercising jurisdiction is consistent with Craig’s Stores itself. Craig’s Stores disclaimed jurisdiction over “post-confirmation claims based on post-confirmation activities.” Enron, 535 F.3d at 325. The suit in Craig’s Stores, 266 F.3d at 389, was brought eighteen months after confirmation—apparently well after the debtor had finished reorganizing. But NFC’s claims concern pre-confirmation activities, and at the time of removal GenOn had not “fully consummated” its plan. U.S. Brass, 301 F.3d at 305. Much work remained to effect GenOn’s plan, prompting the bankruptcy court to express “extreme concerns” that the plan would stall short of “the finish line.” NFC’s claims threatened to destabilize the fragile consensus around a settlement crucial to the success of GenOn’s plan. There thus was no question that, when GenMa removed, those claims might hinder the plan’s “completion.” Ibid. Such “attenuated, hypothetical effects of third-party litigation can give rise to related-to bankruptcy jurisdiction.” Fire Eagle, L.L.C. v. Bischoff (In re Spillman Dev. Grp.), 710 F.3d 299, 305 (5th Cir. 2013). Related-to jurisdiction existed at removal. 2. NFC objects that there is no evidence to support GenMa’s assertion of related-to jurisdiction. It also warns that our holding will extend related-to jurisdiction to every controversy under the sun. We disagree on both counts. a. We start with the evidentiary objections. The first is that GenMa “offered no evidence that NFC’s lawsuit” could affect GenOn’s plan. The second is that there is contrary evidence: GenMa’s notice of removal related that GenOn expected to complete its bankruptcy proceedings soon. And that notice did not say that GenOn could not reorganize unless NFC’s claims were resolved. NFC also points out that GenOn ultimately exited bankruptcy. NFC relies on those facts to show that “[n]othing in the plan turned on the results of this case.” We dismiss NFC’s first objection. Though neither the bankruptcy court nor the Southern District of New York made detailed findings of fact,[12] the record shows that NFC’s claims imperiled GenMa’s ability to perform the settlement. NFC does not dispute the terms of the proposed settlement, which required GenMa to maintain cash reserves and to pay off the Lessors’ debts on the leases. Nor could NFC contest that the framework for that settlement became part of GenOn’s plan, that the bankruptcy court deemed that settlement essential to GenOn’s restructuring, or that NFC had asserted tens of millions in damages against GenMa at the time of removal. Those damages could have threatened GenMa’s ability to perform the settlement: GenMa’s financial filings had expressed “substantial doubt” of its own ability to avoid bankruptcy as late as December 2017, when the settlement framework became part of the plan. We thus do not doubt that “this proceeding” could have affected “compliance with or completion of” GenOn’s plan. U.S. Brass, 301 F.3d at 305. NFC’s second objection—that the evidence belies the relatedness of its claims—is unavailing. It is true that GenMa’s removal notice did not say that NFC’s claims would torpedo GenOn’s reorganization or that GenOn’s plan could not progress unless those claims were resolved. But GenMa did not have to say that. To meet the test for post-confirmation jurisdiction, GenMa just had to show that the dispute “pertain[ed] to the plan’s implementation or execution.” Ibid. (emphasis added). A threat to the plan, even if contingent, supported removal. The fact that GenOn eventually exited bankruptcy does not change that conclusion: “If related-to jurisdiction actually existed at the time of removal,” later events “cannot divest the district court of that subject-matter jurisdiction.” Enron, 535 F.3d at 336 (cleaned up). NFC’s evidentiary objections are unpersuasive. b. NFC also warns that ruling for GenMa would swell related-to jurisdiction well beyond its proper limits. That contention deserves careful attention. In its briefing, GenMa asserts other reasons for jurisdiction besides the settlement enshrined in GenOn’s plan. GenMa points out, for example, that NFC’s claims reduce GenMa’s value and thus “diminish[ ] GenOn’s enterprise value.” And by doing that, GenMa continues, NFC’s claims would reduce the “collateral available [to GenOn] to obtain . . . exit financing” in the reorganization. But we agree with NFC: “If such an indirect impact” on the debtor “could establish jurisdiction, every lawsuit against any debtor’s subsidiary (direct or indirect) would be swept into bankruptcy court—indeed, any litigation affecting any debtor’s assets would be related to’ the bankruptcy.” But that is not our rule.[13] Instead, our rule is that post-confirmation jurisdiction is proper only where the dispute pertains to the plan’s implementation or execution. Few disputes between non-debtors qualify. Neither “judicial economy” nor “[s]hared facts between the third-party action” and a bankruptcy conflict can alone sustain that jurisdiction. Zale, 62 F.3d at 753. To fall within our post-confirmation jurisdiction, a dispute typically must implicate a specific plan’s provision[14] or the parties’ bankruptcy-law rights or responsibilities.[15] Moreover, we have often found jurisdiction only after observing that the parties’ post-confirmation dispute “principally dealt with [pre]-confirmation relations between the parties,” thus satisfying the first factor from Craig’s Stores, 266 F.3d at 391. E.g., Galaz, 841 F.3d at 322; Enron, 535 F.3d at 335–36. Those strictures are many and meaningful, but none precludes our power here. The text and structure of Congress’s jurisdictional grant confirm that conclusion. See 28 U.S.C. § 1334(b). Its “abstention provisions” suggest that we may not unduly narrow our jurisdictional gaze. Wood v. Wood (In re Wood), 825 F.2d 90, 93 (5th Cir. 1987). “Congress wisely chose a broad jurisdictional grant and . . . broad abstention doctrine[s] over a narrower jurisdictional grant so that the district court could determine in each individual case whether hearing it would promote or impair efficient and fair adjudication of bankruptcy cases.”[16] Jurisdiction existed, and removal was proper. III. Having established the district court’s jurisdiction, we now must decide whether the court had to abstain from its exercise. We agree with the district court: Section 1334(c)(2) did not require abstention. Section 1334(c)(2) requires courts with bankruptcy jurisdiction to abstain from hearing “State law claim[s] or State law cause[s] of action” where four conditions are met: First, “an action” “with respect to” those state-law claims “could not have been commenced” in federal court absent bankruptcy jurisdiction. Ibid. Second, the claims only relate to a bankruptcy case; that is, they are not core bankruptcy claims. Ibid. Third, an action regarding the claims “has been commenced in state court” and, fourth, such an action can be “timely adjudicated” there. TXNB, 483 F.3d at 300 (quoting Schuster v. Mims (In re Rupp & Bowman), 109 F.3d 237, 239 (5th Cir. 1997)). The statute requires abstention only if all four conditions are met. We review for abuse of discretion a district court’s refusal to abstain. Id. at 299. No one disputes that the second and third requisites are met: NFC’s contract claims “relate [only] to” GenOn’s bankruptcy case, and those claims began in state court before GenMa removed them. The parties dispute only the first and fourth requisites: whether NFC’s claims “could not have been commenced” in federal court absent bankruptcy jurisdiction, and whether a state court could have “timely adjudicated” those claims. 28 U.S.C. § 1334(c)(2). But we need only address the first requisite here. It was not met, so we may proceed to the merits. Mandatory abstention applies only if “an action” “with respect to” the claims that are “related to” the bankruptcy proceeding “could not have been commenced” in federal court “absent” bankruptcy jurisdiction. Ibid. That requisite is easy to apply where an independent basis for federal jurisdiction exists from the start. See, e.g., TXNB, 483 F.3d at 299–301. But none then existed here. GenMa could not have removed NFC’s claims to federal court on any ground other than related-to bankruptcy jurisdiction. That’s because “federal diversity-of-citizenship jurisdiction depends on the state of things at the time of the action brought.’”[17] Our parties are diverse[18] and contest tens of millions of dollars. But when GenMa removed NFC’s claims, there were parties in NFC’s state-court suit—namely, the indenture trustee and the Lessors—whose presence precluded diversity jurisdiction. This court can forgive a want of diversity at removal in cases meeting the requisites of diversity jurisdiction when judgment was entered. Caterpillar Inc. v. Lewis, 519 U.S. 61, 75 (1996). Yet that mercy is limited to cases “tried on the merits,” which do not include cases resolved on motions for dismissal or summary judgment. Camsoft Data Sys., Inc. v. S. Elecs. Supply, Inc., 756 F.3d 327, 337 (5th Cir. 2014). NFC says that principle ends this case. Not so fast. Though it’s true that diversity jurisdiction did not support removal of NFC’s claims, that’s not what Section 1334(c)(2) asks. That statute commands abstention only where “an action” regarding the claims before the federal district court “could not have been commenced” in a federal court absent bankruptcy jurisdiction. In other words, federal courts must abstain only if “the claim” in the federal court “ has no independent basis for federal jurisdiction.” TXNB, 483 F.3d at 300 (emphasis added and alteration adopted). The statute thus states a two-part inquiry. The first question is which “State law claim[s] or State law cause[s] of action” were properly before the district court. 28 U.S.C. § 1334(c)(2). Here, only NFC’s claims against GenMa were before that court. The claims against the nondiverse parties stayed in state court, thanks to the bankruptcy removal statute. Unlike the general removal statute, which allows a defendant to remove only a “civil action,” § 1446(a) (emphasis added), the bankruptcy removal statute permits removal of “any claim or cause of action in a civil action” to the federal court in that district, § 1452(a) (emphasis added). That’s what GenMa did. It removed under Section 1452, bringing only the state-law claims against it to the Southern District of New York. The statute next asks whether “an action” regarding those claims— the claims before the federal court—”could . . . have been commenced” in federal court absent related-to jurisdiction. § 1334(c)(2). Here, the answer is yes. Diversity jurisdiction would exist over those claims, see § 1332, because the parties dispute more than $75,000 and because NFC and GenMa are citizens of different states. So abstention was not required. NFC says that that reading of the statute is “baseless.” According to NFC, the statute’s “commencement requirements” address “the action filed in state court,” not the “embedded State-law claims’ related to’ the bankruptcy.” (Alterations adopted) (quoting § 1334(c)(2)). And because NFC’s state-court action “indisputably involved nondiverse parties,” NFC deems this abstention factor satisfied. NFC misreads the statute. Here is what the full subsection says: Upon timely motion of a party in a proceeding based upon a State law claim or State law cause of action, related to a case under title 11 but not arising under title 11 or arising in a case under title 11, with respect to which an action could not have been commenced in a court of the United States absent jurisdiction under this section, the district court shall abstain from hearing such proceeding if an action is commenced, and can be timely adjudicated, in a State forum of appropriate jurisdiction. § 1334(c)(2) (emphasis added). That text does not permit NFC’s reading. The phrase “with respect to which” links “an action”—the one that interests us—to the “claim[s]” and “ cause[s] of action” on which the federal “proceeding” is “based.” And when the statute finally mentions the “action . . . commenced . . . in a State forum,” it refers to that action as “an action”—not as “that action,” “such action,” or other words suggesting that the action at the statute’s end (the one “commenced in a State forum”) is the same action mentioned at its start (the one that “could not have been commenced” in federal court). Read plainly, the statute asks whether an actionany action with respect to NFC’s state-law claims against GenMa could have been commenced in a federal court. Ibid. Because the answer is yes, abstention was not required. Besides its erroneous reading of the statute, NFC invokes two out-ofcircuit district court cases[19] for the notion that GenMa’s failure “to list diversity jurisdiction in its notice of removal” bars GenMa from citing diversity jurisdiction as a reason not to abstain. That reading is wrong. It is true that a notice of removal must contain “a short and plain statement of the grounds for removal.” 28 U.S.C. § 1446(a). But a removal notice concerns the basis for our jurisdiction. Mandatory abstention does not. It is not jurisdictional. It is forfeitable, and it cannot apply absent a “timely motion” of the party wishing to avoid the federal forum. § 1334(c)(2); see also VSP Labs, Inc. v. Hillair Cap. Invs., L.P. (In re PFO Global, Inc.), 26 F.4th 245, 254 (5th Cir. 2022). GenMa thus had no reason to address abstention in its removal notice before NFC asked for it. “We do not require a litigant to anticipatorily rebut all potential arguments his adversary may raise.” Hoyt v. Lane Constr. Corp., 927 F.3d 287, 296 n.2 (5th Cir. 2019). We find no abuse of discretion. Abstention was not required. IV. At last, we turn to the merits. NFC lost all four of its claims before the bankruptcy court. It then lost at the district court, which, taking the bankruptcy court’s advice, disposed of all claims and entered a take-nothing judgment against NFC. NFC now appeals the Rule 12(b)(6) dismissal of its implied-covenant and reformation claims and the summary judgments dismissing its claims for breach of warranty and indemnification. NFC cannot prevail, so we affirm the district court in all respects. A. We begin with the summary judgments. The district court granted those judgments to GenMa on NFC’s breach-of-warranty and indemnification claims. NFC asks us to reverse both judgments. Summary judgment is proper if there’s no genuine dispute over the meaning of the parties’ contract, as when that contract is unambiguous. Tekelec, Inc. v. Verint Sys., Inc., 708 F.3d 658, 664 (5th Cir. 2013).[20] So we may affirm the summary judgments if the relevant contracts are clear that GenMa breached no warranty and owes no indemnity. And those contracts are clear as day. 1. Here is the breach-of-warranty theory: By paying NFC to procure letters of credit from Natixis, GenMa created a lien on its assets, breaking two promises to the Lessors.[21] And that deal breached GenMa’s warranty to NFC not to break those promises.[22] That breach of warranty was a default under the parties’ Payment Agreement. That theory rests on the premise that GenMa’s paying NFC created a “Lien”—namely, a proscribed “deposit arrangement.” If the payment effected a lien on GenMa’s assets, NFC posits, then GenMa broke its promises to the Lessors not to do that and to maintain uncollateralized credit support. But that premise is wrong. GenMa’s leases define “Lien” as “any security interest, security deed, mortgage, pledge, hypothecation, assignment, deposit arrangement, encumbrance, lien (statutory or otherwise), lease, title retention arrangement, charge against or interest in property . . . to secure payment of a debt or performance of an obligation.” That definition is prolix and redundant, as only a lawyer could love. But it is simpler than it appears. Each of the proscribed arrangements conveys to a creditor a conditional interest in the debtor’s property to secure an obligation. In other words, each resembles a conventional lien.[23] A good example is the classic pledge: Needing a loan, a beggar pawns his watch. The pawnbroker gives the loan but keeps the watch. The watch still belongs to the beggar. If he repays the loan, his watch is returned. If he does not repay, the pawnbroker may sell the watch to cover the debt—but he owes the beggar any surplus proceeds. See U.C.C. § 9-615(d)(1). It is, after all, his watch. Or consider a common deposit arrangement: To rent an apartment, a tenant deposits $1,000 with his landlord at the start of the lease. If, at the end of the lease, the tenant has kept the place tidy, the landlord must return the deposit. But if the tenant puts holes in the walls, stops paying rent, or otherwise breaches the lease, the landlord may apply the deposit “to cover [her] losses or damages.”[24] The Payment Agreement is nothing like those arrangements: GenMa unconditionally conveyed the $130 million payment to NFC, and that payment did not secure any duty of GenMa to NFC or to anyone else. GenMa surrendered all interest in its payment to NFC. On that point, the Agreement could not be clearer. It expressly and emphatically disclaims any interest of GenMa in the funds it paid NFC. The Agreement says those funds “are and shall become the property of NFC.” It also says that “[o]nce [those funds are] paid, [GenMa] will have no interest, claim, reversionary or residual interest” in them. (Emphasis added.) The Agreement commits GenMa “ to irrevocably pay in full to NFC the Payment Amount . . . to procure [NFC's] services.” (Emphasis added.) The parties even solemnize that “the Payment Amount has been indefeasibly paid by [GenMa] to NFC,” that NFC “bears the risk of loss on the Payment Amount,” and that NFC can keep “any returns, interest, gains or other earnings that accrue on the Payment Amount.” For its payment to effect a lien on its assets, GenMa had to retain some ongoing interest in the funds or some right to their return. But the Payment Agreement repudiated all such rights—not once, not twice, but throughout. But even if GenMa had retained some theoretical interest in its payment to NFC, that payment secured no duty or obligation of GenMa to NFC or to anyone else. Paying NFC $130 million and the letter-of-credit fee was GenMa’s only “obligation” under the Agreement. In exchange, NFC promised to acquire letters of credit and, if they went undrawn, to pay rent to the Lessors. Those duties fell squarely and only on NFC. “NFC shall be solely responsible” for satisfying any draws on the letters of credit, the Agreement states, and “neither [GenMa] nor any Affiliate thereof shall have any liability for, and no claim or recourse against [GenMa] or any Affiliate thereof will exist with respect to, any reimbursement obligation” arising from such draws. That arrangement looks nothing like a lien, where specific assets are set aside to secure some obligation of the debtor. See, e.g., Biloxi Casino Belle, 368 F.3d at 493–94. Because GenMa gave up any interest in the $130 million paid to NFC, and because those funds did not secure any of GenMa’s debts or obligations, the payment effected “no lien or other charge . . . against GenMa’s assets.” And absent a lien on its own assets, GenMa did not breach its warranty that the Agreement complied with the leases. Against those points, NFC first urges that affirmance would ignore the true “economic substance” of the parties’ transaction and instead credit the contract’s “artificial labeling.” (Quotation omitted.) According to NFC, GenMa’s payment was a “clear” deposit arrangement. Not so. Though courts should take care not to credit nomenclature over reality, here it is NFC pressing unreality. Any inquiry into economic substance must rest on what the parties agreed, see Int’l Trade Admin. v. Rensselaer Polytechnic Inst., 936 F.2d 744, 749 (2d Cir. 1991), and GenMa disclaimed any right to the funds it paid NFC. Nothing in the contract supports NFC’s insinuation that those statements were a sham. GenMa paid NFC $130 million for its unsecured promise to provide credit support or to pay rent—nothing more. It makes no difference that NFC agreed to procure letters of credit to satisfy GenMa’s credit-support obligations and to apply any unused funds to GenMa’s rent. Those promises benefited GenMa; that is why the parties contracted. What’s key is that neither promise preserved for GenMa any interest in the funds it paid NFC. To the contrary, NFC’s promises are what those funds were irrevocably conveyed to purchase. There was no lien. The endpoint of NFC’s theory reveals its illogic. At bottom, NFC asks this court to conclude that the Agreement effected a lien—despite all textual evidence to the contrary—because GenMa exchanged $130 million for promises of equal value.[25] NFC stresses that GenMa’s payment equaled the value that NFC had committed to provide between the letters of credit and the rent payments, and that the Agreement “dictated” the terms of those return promises. But those facts also would describe any purchase. NFC’s theory would transform every purchase of any promise, however mundane, into a secured transaction. That is not the law. Paying someone to stock the office vending machine is not a secured transaction—even if the vendor agrees to stock candies that are worth what was paid[26] and even if the candies’ size, color, and flavor are “dictated” by contract.[27] NFC also faults the district court for “ignor[ing] . . . central evidence” that GenMa treated their transaction as a deposit and, thus, a proscribed lien. According to NFC, GenMa accounted for the transaction as a long-term deposit and disclosed it as one in its SEC filings. NFC also cites testimony purporting to exhume the parties’ “shared understanding” of the leases’ collateralization requirement and bar on liens. But it is blackletter law that extrinsic evidence is “inadmissible to add to or vary the writing” when that writing is unambiguous. W.W.W. Assocs., 566 N.E.2d at 642. NFC offers no reason for this panel to disregard that “familiar and eminently sensible proposition of law,” ibid., so we will not. The Agreement created no lien, so we affirm the summary judgment for GenMa on this claim. 2. NFC next asserts that the Agreement requires GenMa to cover NFC’s costs in this suit and in various state-court actions between NFC and the Lessors. We disagree and affirm the judgment for GenMa on that claim. NFC claims two indemnities in the Agreement. The first does not apply. It is an indemnity for costs incurred to enforce NFC’s rights under the Agreement. But it kicks in only if GenMa defaults, and the only asserted default is GenMa’s breach of warranty. GenMa breached no warranty, so it need not cover NFC’s costs in this suit. The second indemnity protects NFC and Natixis against all losses “in connection with any . . . judicial proceeding . . . brought or threatened (by any third party, by [GenMa] or any Affiliate [of GenMa]) arising out of or relating in any respect to this Agreement or any Letter of Credit.” Though broad, that indemnity does not cover “any reimbursement obligation to [Natixis] or any other Person in respect of any” lease payments or disbursements under the letters of credit. NFC says that second indemnity covers its losses in two sets of cases. The first comprises the sundry suits by the Lessors against Natixis for failing to honor their draws on the Natixis letters of credit. The second is this suit, as it proceeded in state court against the Lessors and the indenture trustee. NFC and Natixis lost both.[28] The Agreement’s indemnity does not cover NFC’s losses from the suits against Natixis. The indemnity bars “any reimbursement obligation to [Natixis] or any other Person in respect of any [letter-of-credit] Disbursement.” (Emphasis added.) The Lessors sued to force Natixis to honor the draws on the letters of credit. When Natixis honors a draw, the Payment Agreement requires NFC to reimburse it. And that reimbursement, NFC candidly admits, is the only loss it suffered. “Although the [Lessors] sued Natixis,” NFC explains in its opening brief, “NFC incurred losses from those suits via its reimbursement obligations under the [Payment A]greement.” (Emphasis added.) That point is conclusive. The suits concerned Natixis’s disbursements under the letters of credit, and NFC’s costs arose from its duty, under the Agreement, to reimburse Natixis for those draws.[29] The indemnity does not cover that. Nor does it cover NFC’s costs in its suit against the Lessors and the indenture trustee. The indemnity covers all losses incurred “in connection with . . . judicial proceeding[s]“—but only those “brought or threatened [ ]by any third party.” Suits that NFC commenced are not “brought or threatened” “by a third party,” so they are excluded from the indemnity. NFC asserts that it sued the Lessors “in response to the[ir] litigation threats.” But the indemnity does not cover suits responding to proceedings “brought or threatened” by third parties; it covers only those proceedings. That plain textual reading makes good sense: Indemnities exist to “protect or preserve against . . . loss.”[30] They generally are not designed to finance offensive litigation. And when they are put to that purpose, the parties leave no doubt—as with the first indemnity provision we discussed, which, as explained, does not apply here.[31] We affirm the summary judgment on the indemnification claim. B. NFC also appeals the Rule 12(b)(6) dismissal of its implied-covenant and reformation claims. Because those claims come to us on a motion to dismiss, we confine our review to the pleadings, their attachments, and any documents the pleadings mention that are central to the plaintiff’s claims. Collins v. Morgan Stanley Dean Witter, 224 F.3d 496, 498–99 (5th Cir. 2000). The pleading standard is familiar: To withstand a motion to dismiss under Rule 12(b)(6), a complaint must present enough facts to state a plausible claim to relief. A plaintiff need not provide exhaustive detail to avoid dismissal, but the pleaded facts must allow a reasonable inference that the plaintiff should prevail. Facts that only conceivably give rise to relief don’t suffice. Thus, though we generally take as true what a complaint alleges, we do not credit a complaint’s legal conclusions or threadbare recitals of the elements of a cause of action. Mandawala v. Ne. Baptist Hosp., 16 F.4th 1144, 1150 (5th Cir. 2021) (cleaned up). Our review is de novo. Ibid. Applying that standard and the contract law of New York, we affirm both dismissals. 1. NFC first contends that GenMa breached the implied covenant of good faith and fair dealing. Even taking as true NFC’s well-pleaded factual assertions, as we must, NFC does not come close to stating that claim. By default, every contract includes a covenant of good faith and fair dealing. Dalton v. Educ. Testing Serv., 663 N.E.2d 289, 291 (N.Y. 1995). That covenant bars “opportunistic behavior” in the gaps of the contract; a classic example is an insurer’s “paying a person who has sued the insured to convert his claim to one not covered by the insurance policy.” In re Ocwen Loan Serv., L.L.C. Mort. Servicing Litig., 491 F.3d 638, 645–46 (7th Cir. 2007) (Posner, J.). Contract law forbids such conduct, even if the contract’s text does not address or anticipate it, because it “destroy[s] or injur[es] the right of the other party to receive the fruits of the contract.” Dalton, 663 N.E.2d at 291 (quotation omitted). Those descriptions of the implied covenant belie its strictures. Courts may not read into a contract a covenant that doesn’t follow inexorably from its terms. See id. at 292.[32] Nor may they rewrite a contract to include terms that a party, with the benefit of hindsight, would have or should have included. Instead, courts may imply only those promises that “a reasonable person in the position of the promisee would be justified in understanding were included” in the bargain from the start. Dalton, 663 N.E.2d at 291 (emphasis added and quotation omitted). According to NFC, the implied covenant here is an implied promise by GenMa not to miss any rent payments to the Lessors. GenMa broke that promise, NFC says, by refusing to pay its rent despite having the funds to do so. That caused the Lessors to draw on Natixis’s letters of credit, which, in turn, deprived NFC of “the primary benefit of [the] bargain”—the chance to earn interest on GenMa’s $130 million payment. We agree with the bankruptcy court: That claim is “baseless.” As the bankruptcy court explained, the “entire purpose” of a credit-support arrangement like this one is to “insure” the beneficiaries “against the risk” of default. The Lessors insisted on letters of credit so that they could draw on them if GenMa didn’t pay its rent. It defies logic to say that the risk of immediate default was “squarely at odds with the parties’ agreement and shared expectations,” where NFC agreed to insure the Lessors against any default. NFC’s desired covenant also contradicts the Agreement’s plain terms. NFC agreed to bear “sole[ ] responsib[ility] for satisfying any reimbursement obligation” and to release GenMa from “any liability for . . . [or] with respect to” any such obligation. NFC also agreed to “bear[ ] the risk of loss on the Payment Amount” and to “provide cash collateral” to back Natixis’s letters of credit. A covenant not to miss rent thus would create liability for GenMa that the Agreement assigned only to NFC. Yet, under New York law, “no obligation can be implied” that would conflict with the contract’s terms. Dalton, 663 N.E.2d at 292. The bankruptcy court cited those reasons and one other: the contract’s integration clause. That clause states that the Payment Agreement is “the entire agreement among the parties relating to the subject matter hereof and supersede[s] all oral statements and prior writings with respect thereto.” NFC faults the court for relying on that clause. Though we share that critique,[33] it does not matter. Waived or not, that duty could not require GenMa to continue paying rent. Managing that risk was the Agreement’s purpose, and the Agreement placed that risk—all of it—on NFC only. We affirm the dismissal of NFC’s implied-covenant claim. 2. We last consider NFC’s reformation claim, which could proceed only if NFC had pleaded both a mutual mistake and the precise terms of the parties’ actual agreement. Chimart Assocs. v. Paul, 489 N.E.2d 231, 234 (N.Y. 1986). We agree with the bankruptcy court: NFC did not adequately plead a mutual mistake. In New York, “the thrust of a reformation claim is that a writing does not set forth the actual agreement of the parties.” Ibid. In other words, reformation requires a mistake “in the drafting of the instrument, not in the making of the contract.”[34] Thus, to survive GenMa’s motion to dismiss, NFC must plead not just (1) a mutual mistake but also (2) “exactly what was really agreed upon between the parties.” George Backer, 385 N.E.2d at 1066. That burden is great. NFC must overcome the “heavy presumption that a deliberately prepared and executed written instrument manifests the true intention of the parties.”[35] In its latest complaint, NFC observes that the parties agreed to cap its liability at $130 million—the amount of GenMa’s payment. But because of a “drafting error,” NFC says, the Payment Agreement omitted that cap, allowing “aggregate draws” on the letters of credit to exceed $130 million. That omission was a mutual mistake, which NFC contends this court should correct by reforming the contract to require GenMa to pay NFC another $45 million—the amount by which the Lessors’ draws exceeded $130 million. But the Agreement contradicts that allegation of mistake.[36] It did cap NFC’s exposure to the letters of credit at $130 million, and it did so with the defined term, “Excess Capacity of Lessor.” Here is how: The “Excess Capacity” of each Lessor equaled that Lessor’s share of GenMa’s $130 million payment minus the sum of four other amounts: (1) the credit support that each Lessor would require for that lease period; (2) the draws not yet paid to that Lessor; (3) the draws already paid to that Lessor; and (4) the total amount of lease payments already made by NFC to that Lessor. To make things more concrete, we can suppose that a Lessor’s share of the payment amount was $100. If that Lessor had drawn $25 on a Natixis letter of credit in a prior period and needed $40 in credit support for the current period, that Lessor’s Excess Capacity would equal $35. If the Lessor then requested another $35 in draws, its Capacity would drop to zero. The Agreement employs the Excess Capacity term to cap what NFC must pay to any Lessor. Any draw or lease payment to a Lessor reduces that Lessor’s Excess Capacity. And when that Excess Capacity drops to or below zero, “NFC shall have no obligation to cause a Letter of Credit to be issued or to remain outstanding.” Likewise, “if a Letter of Credit is drawn, the outstanding amount of the applicable Letter of Credit and future outstanding amounts available to be drawn thereunder . . . shall be permanently reduced by the amount of such draw.” The result, the Agreement confirms, is that “no Letters of Credit will need to be provided to any Lessor if there is not sufficient Excess Capacity . . . available” to that Lessor. Confirming those statements, the Agreement details exactly how much credit support each Lessor must get in each period, absent letter-of-credit draws. And in no period does the total required credit support exceed $130 million. With those draw caps staring it in the face, NFC tries to conflate the Payment Agreement with the letters of credit that its affiliate issued. The form letter of credit attached to the Agreement did not include draw caps, and that form, NFC says, dictated the form of the letters that Natixis issued. And, because NFC could not have added to the letters a provision capping its liability at $130 million, NFC contends, any mistake in those letters was mutual—attributable to NFC and GenMa. The form letter of credit did not bar Natixis from adding draw caps to the letters it issued. But even if we were wrong about that, Natixis could have prevented excessive draws in ways that would have complied with the form letter of credit and the Agreement. As NFC acknowledges, the Agreement says that NFC would never have to backstop more than $130 million in credit support in a single period. NFC highlights that the excessive draws occurred because the parties did not account for “the highest possible draw . . . across multiple periods.” But nothing prevented Natixis from posting letters of credit for each lease period and each Lessor, so that it could adjust future letters to reflect that Lessor’s remaining capacity. Nor did the form letter of credit preclude posting all letters for all periods at the same time on the condition that no Lessor could draw more than its share of the $130 million. The schedule to the form letter of credit lists no specific dollar amounts, and the form letter leaves blank its expiration date; those spaces are left for Natixis to fill.[37] Thus, even if the form letter of credit precludes draw caps, as NFC asserts, NFC and Natixis could have guarded against the risk of excessive draws by issuing letters of credit period by period. Instead, Natixis issued every letter of credit at once, exposing itself to far more than $130 million in liability. Each letter “irrevocably authorize[d]” the Lessor “ to draw on [Natixis] . . . in any amount up to an aggregate amount as of any date within the applicable period as set forth on Schedule I . . . .” That schedule listed all the credit-support amounts that the Agreement required Natixis to post for that Lessor through the year 2030. And because the letters did not cap draws across periods, Natixis posted credit support far exceeding each Lessor’s share of the $130 million payment. All told, “Natixis posted $2.2 billion in letters of credit, covering every remaining rent period for over a decade.” (Emphasis added.) Nothing in the Agreement or the form letter of credit compelled that colossal, unilateral blunder. And absent fraud, which is not alleged here, a unilateral mistake cannot support reformation. Chimart, 489 N.E.2d at 234. NFC lobs one more argument: Had it capped the letters of credit or issued letters only to the extent of each Lessor’s share of GenMa’s $130 million payment, GenMa would have defaulted on its leases. That is because, in some periods, the credit support required by the leases for a given Lessor exceeded that Lessor’s share of GenMa’s $130 million payment. For example, in one lease period, one Lessor needed $57.5 million in credit support, but that amount exceeded its share of GenMa’s payment by about $1.7 million. NFC says that result would make no sense; after all, the point of the Agreement was to satisfy GenMa’s lease obligations. Why should NFC bear sole responsibility for “fail[ing] to identify and ameliorate [that] mutual mistake”? There is some truth in that contention. The parties may have miscalculated the amount of credit support needed to satisfy GenMa’s lease obligations. But that mistake, mutual or not, was GenMa’s problem. Had Natixis carefully crafted its letters of credit, NFC would not have had to pay any more to the Lessors than GenMa had paid it, no matter how badly the parties miscalculated the credit support that GenMa’s leases required. We agree with GenMa: “NFC cannot demand more money from GenMa for discovering that it could have obtained less credit support” than the Agreement required. Natixis made a mistake that cost NFC tens of millions of dollars. But reformation cannot erase that unforced blunder. The dismissal was correct. The judgment is in all respects AFFIRMED.
