We must decide whether the district court erred by refusing to remand this case to state court. It did not. Next, we must decide whether the district court erred by granting summary judgment to the defendant. It did.
I.
On December 29, 2015, Jeffery Hoyt hit a patch of ice while driving on FM 2264 in Wise County, Texas. Jeffery slid off the road. His car landed upside down in an adjacent body of water. Tragically, Jeffery drowned. Less than an hour later, a second driver hit the same patch of ice. The second driver likewise slid off the road. And the second driver landed directly on top of Jeffery's submerged vehicle. That apparently saved the second driver from drowning. First responders rescued him and, in the process, discovered Jeffery's vehicle and body.
On September 20, 2016, members of Jeffery's family ("the Hoyts") filed suit in Texas state court. They sued C.E.N. Concrete Construction Co., Storm Water Management, Inc., and the Lane Construction Corporation. The Hoyts, C.E.N., and Storm are citizens of Texas. Lane is not. The Hoyts contended all three companies had performed construction work on FM 2264 and caused ice to form at the crash site. The defendants moved for summary judgment. The state district court granted C.E.N.'s motion and entered a "take nothing" judgment in its favor.
The Hoyts and Storm engaged in settlement discussions. They never reached *292agreement. Yet on September 22, 2017-one year and two days after the suit began-the Hoyts voluntarily dismissed their claims against Storm. The Hoyts received no compensation from Storm.
Five days later, Lane removed the case to federal court on the theory that it now fit within federal diversity jurisdiction. See 28 U.S.C. § 1332(a)(1) ; Lincoln Prop. Co. v. Roche , 546 U.S. 81, 89, 126 S.Ct. 606, 163 L.Ed.2d 415 (2005) (requiring "complete diversity"). The next day, the Hoyts filed an emergency motion to remand. They argued Lane's notice of removal was untimely. The federal district court denied that motion. In a second motion to remand filed about a month later, the Hoyts argued the voluntary-involuntary rule prohibited removal because C.E.N. had been dismissed against their wishes. The district court denied that motion too.
Lane moved for summary judgment on the Hoyts' claims for premises liability and gross negligence. The federal district court granted the motion. It dismissed the claims against Lane with prejudice. The Hoyts timely appealed.
II.
The Hoyts argue we must remand the case to state court. We disagree. The district court properly rejected both remand motions.
A.
The Hoyts' first motion for remand turns on timeliness. Under 28 U.S.C. § 1446(c), the defendant in a diversity case has one year following the commencement of an action to remove it. But Congress created an exception to this time bar where "the district court finds that the plaintiff has acted in bad faith in order to prevent a defendant from removing the action." Id. § 1446(c)(1). Here, the district court found the Hoyts acted in bad faith by improperly joining Storm (which prevented complete diversity and hence precluded removal).1 The district court therefore denied the Hoyts' motion to remand under § 1446(c) 's time bar.
Although we review the denial of a motion to remand de novo , we review the underlying finding of bad faith for clear error. Spear Mktg., Inc. v. BancorpSouth Bank , 791 F.3d 586, 591 (5th Cir. 2015). We hold (1) the district court's bad-faith finding was not clearly erroneous, and (2) the Hoyts cannot avoid that result by relying on cases that predate Congress's enactment of the bad-faith exception to § 1446(c)(1) 's time bar.
1.
Exercising its role as factfinder, the district court found the Hoyts "knew months beforehand that the evidence would not support the claims against Storm." That was not clear error. The Hoyts dismissed Storm a mere two days after the one-year deadline expired. They did so without receiving any consideration from Storm. Before that dismissal, the Hoyts seem to have pursued their claim against Storm only half-heartedly. Their witness list for trial did not include any fact witnesses from Storm. And the Hoyts' expert witnesses made no serious efforts to establish Storm's liability. All of this suggests the Hoyts kept Storm in the case for one purpose and one purpose only-to prevent *293removal during § 1446(c) 's one-year removal period. Two days after accomplishing that purpose, the Hoyts dismissed Storm for free.
The Hoyts' response is unpersuasive. In the district court, the Hoyts submitted an affidavit from their attorney to describe allegedly strategic reasons for their decision to dismiss Storm. But these explanations relate to why the Hoyts were reluctant to go to trial against Storm or accept Storm's (apparently low) settlement offer. They do not explain why the Hoyts waited until just two days after the one-year deadline to dismiss Storm. And while the Hoyts claim they dismissed Storm after their settlement discussions came to naught, the district court found it "suspicious[ ]" the Hoyts did not clarify when "the alleged discussions with Storm" took place. We agree.
Nor can the Hoyts win a remand by raising "fact issues" regarding their good faith. True, "we resolve all contested factual issues ... in favor of the plaintiff" when determining whether it improperly joined a non-diverse defendant. Gasch v. Hartford Accident & Indem. Co. , 491 F.3d 278, 281 (5th Cir. 2007). That makes sense because, when considering whether "the plaintiff [is able] to establish a cause of action against the non-diverse party in state court," the question is what the plaintiff might prove in the future . Ibid. ; see Guillory v. PPG Indus., Inc. , 434 F.3d 303, 308 (5th Cir. 2005) ; Smallwood v. Ill. Cent. R.R. Co. , 385 F.3d 568, 573 (5th Cir. 2004) (en banc). When it comes to bad faith, by contrast, the question is what motivated the plaintiff in the past -that is, whether the plaintiff's litigation conduct meant "to prevent a defendant from removing the action." 28 U.S.C. § 1446(c)(1). The district court found as a matter of fact the Hoyts acted in bad faith. The Hoyts failed to carry their burden to prove that finding was clearly erroneous.
2.
The Hoyts also argue we must reverse because their litigation conduct does not satisfy the exception we created in Tedford v. Warner-Lambert Co. , 327 F.3d 423 (5th Cir. 2003). But Tedford pre-dates Congress's enactment of the bad-faith exception in § 1446(c)(1), and it therefore does not control.
Before 2011, § 1446 prohibited defendants like Lane from removing a case "more than 1 year after commencement of the action"-full stop. 28 U.S.C. § 1446(b) (2006). The statutory text contained no exceptions. Believing this old version of § 1446(b) was "not inflexible," however, Tedford ruled "the conduct of the parties may affect whether it is equitable to strictly apply the one-year limit." 327 F.3d at 426. In light of the plaintiff's "efforts to manipulate statutory rules," the Tedford Court concluded, "[e]quity demands [the plaintiff] be estopped from seeking to remand the case on the basis of the one-year limit in § 1446(b)." Id. at 428 & n.13.
Other courts disagreed. They held that § 1446 did not allow for equitable tolling or estoppel. See, e.g. , Brock v. Syntex Labs., Inc. , No. 92-5740, 7 F.3d 232, 1993 WL 389946, at *1 (6th Cir. Oct. 1, 1993) ; Kinabrew v. Emco-Wheaton, Inc. , 936 F. Supp. 351, 352 n.1 (M.D. La. 1996) (collecting cases). This split persisted until 2011, when Congress amended § 1446. See Federal Courts Jurisdiction and Venue Clarification Act of 2011, Pub. L. No. 112-63, § 103(b), 125 Stat. 758, 760 (2011). By adding the bad-faith exception to the one-year deadline, Congress resolved the conflict.
If Congress wanted to resolve the conflict by adopting the Tedford standard, it could have done so. Cf.
*29442 U.S.C. § 2000bb(b)(1) (listing the restoration of "the compelling interest test as set forth in Sherbert v. Verner , 374 U.S. 398, 83 S.Ct. 1790, 10 L.Ed.2d 965 (1963) and Wisconsin v. Yoder , 406 U.S. 205, 92 S.Ct. 1526, 32 L.Ed.2d 15 (1972)" as a purpose of the Religious Freedom Restoration Act). But it did not. Congress instead chose to replace Tedford 's equitable-estoppel principle with a "bad faith" standard. See 28 U.S.C. § 1446(c)(1). And we presume that choice of different text carries with it a choice of different meaning. Cf. Henson v. Santander Consumer USA Inc. , --- U.S. ----, 137 S. Ct. 1718, 1723, 198 L.Ed.2d 177 (2017) ("[W]hen we're engaged in the business of interpreting statutes we presume differences in language ... convey differences in meaning."). We therefore no longer apply the old § 1446 and the Tedford exception we created. We now apply the new § 1446 and the bad-faith exception Congress created. See Thompson v. Deutsche Bank Nat'l Tr. Co. , 775 F.3d 298, 303 (5th Cir. 2014).
Our holding is consistent with the principle that "Congress is vested with the power to prescribe the basic procedural scheme under which claims may be heard in federal courts." Patsy v. Bd. of Regents , 457 U.S. 496, 501, 102 S.Ct. 2557, 73 L.Ed.2d 172 (1982). Once Congress has prescribed those procedures, we cannot add to them. See, e.g. , Ross v. Blake , --- U.S. ----, 136 S. Ct. 1850, 1857, 195 L.Ed.2d 117 (2016) (holding the Prison Litigation Reform Act does not admit of judge-made exceptions). We do not exalt judge-made doctrines over valid laws. See Am. Express Co. v. Italian Colors Rest. , 570 U.S. 228, 235-39, 133 S.Ct. 2304, 186 L.Ed.2d 417 (2013) (refusing to apply "a judge-made exception to the FAA" broadly because "[t]he FAA does not sanction such a judicially created superstructure"); CLS Bank Int'l v. Alice Corp. Pty. , 717 F.3d 1269, 1303 (Fed. Cir. 2013) (en banc) (Rader, C.J., concurring in part and dissenting in part) ("[J]udge-made exceptions to properly enacted statutes are to be narrowly construed.").
It does not matter the House Report accompanying the 2011 statute cites Tedford . That's for two reasons. First, when "a statute's text is clear, courts should not resort to legislative history." Adkins v. Silverman , 899 F.3d 395, 403 (5th Cir. 2018). And we think the new § 1446(c)(1) is clear. But second, in all events, the report cites Tedford only once, and only then to explain the conflict the amendment would resolve. See H.R. REP. 112-10, at 15 (2011), as reprinted in 2011 U.S.C.C.A.N. 576, 580. The report does not say the bill would adopt the Tedford standard. Instead, it refers to the statutory bad-faith exception as a "new standard." Ibid. Relying on a single reference to Tedford -without accounting for the author's description of the bad-faith standard as "new"-would be "akin to 'looking over a crowd and picking out your friends.' " Patricia M. Wald, Some Observations on the Use of Legislative History in the 1981 Supreme Court Term , 68 IOWA L. REV. 195, 214 (1983) (quoting Harold Leventhal's description of problems with citing legislative history).
Thus, even if the Hoyts' litigation conduct would not satisfy the old Tedford standard (an issue we need not decide), that is no reason to reverse the district court's finding of "bad faith" under the new § 1446(c)(1).
B.
The Hoyts' second motion to remand turned on the voluntary-involuntary rule. Removal was improper, they said, because diversity jurisdiction existed only after C.E.N. exited the case, and because C.E.N. exited the case pursuant to an opposed *295motion for summary judgment, not a voluntary act of the Hoyts. The Hoyts' argument implicates the federal removal statute, a judge-made exception to the statute, and a judge-made exception to that exception. We explain each in turn.
We start with the only easy part-the statutory text. Section 1446(b)(3) is the provision governing removal of a case that becomes removable after the initial pleading is filed. Subject to exceptions not applicable here:
if the case stated by the initial pleading is not removable, a notice of removal may be filed within 30 days after receipt by the defendant ... of a copy of an amended pleading, motion, order or other paper from which it may first be ascertained that the case is one which is or has become removable.
28 U.S.C. § 1446(b)(3).
One might think a case is "removable" under § 1446 so long as it "may be removed" under § 1441. Indeed, courts generally understand § 1441 as defining removability. See Teamsters Local 404 Health Servs. & Ins. Plan v. King Pharm., Inc. , 906 F.3d 260, 266 (2d Cir. 2018) ("[D]efin[ing] what kind of proceedings are removable ... is the ambit of § 1441."); Medlin v. Boeing Vertol Co. , 620 F.2d 957, 964 (3d Cir. 1980) ("Section 1441(b) ... defin[es] removability ...."). And with exceptions not relevant here, § 1441 provides "any civil action brought in a State court of which the district courts of the United States have original jurisdiction, may be removed." 28 U.S.C. § 1441(a).
But this is an area thick with binding precedent. And precedent tells us removability under §§ 1441 and 1446 is subject to a judge-made exception: "[W]here the case is not removable because of joinder of defendants," only "the voluntary dismissal or nonsuit by [the plaintiff] of a party or of parties defendant" can convert a nonremovable case into a removable one. Great N. Ry. Co. v. Alexander , 246 U.S. 276, 281, 38 S.Ct. 237, 62 L.Ed. 713 (1918). We have described this "judicially-created voluntary-involuntary rule" as providing "an action nonremovable when commenced may become removable thereafter only by the voluntary act of the plaintiff." Crockett v. R.J. Reynolds Tobacco Co. , 436 F.3d 529, 532 (5th Cir. 2006) (quotation omitted).
The Supreme Court developed this principle before § 1446 expressly provided for removal of cases that were not initially removable. But in 1949, Congress amended § 1446(b) to allow removal "after receipt ... of an amended pleading, motion, order or other paper from which it may first be ascertained that the case is ... removable." Act of May 24, 1949, Pub. L. No. 81-72, § 83(a), 63 Stat. 89, 101 (1949). Then we held the new § 1446(b) did not abrogate the voluntary-involuntary rule. See Weems v. Louis Dreyfus Corp. , 380 F.2d 545, 547-48 (5th Cir. 1967).
The judicially created voluntary-involuntary rule is itself subject to a judicially created exception for improper joinder. See Great N. Ry. , 246 U.S. at 282, 38 S.Ct. 237 (recognizing the voluntary-involuntary rule applies only "in the absence of a fraudulent purpose to defeat removal"). When the non-diverse defendant was improperly joined, "the voluntary-involuntary rule is inapplicable." Crockett , 436 F.3d at 532.
Beneath this doctrinal thicket lurks a question presented-namely, whether C.E.N. was improperly joined.2
*296We hold it was. Therefore, the voluntary-involuntary rule does not apply. And because the voluntary-involuntary rule does not apply, the case was removable under §§ 1441 and 1446.
Improper joinder "can be established by demonstrating," among other things, the "inability of the plaintiff to establish a cause of action against the non-diverse party in state court." Crockett , 436 F.3d at 532 (quotation omitted). Assessing a plaintiff's ability to establish a cause of action against a non-diverse defendant in state court turns on whether the state courts have already ruled on the merits of the plaintiff's claim. If a state court has not yet ruled, the federal court must ask whether "there [is] any reasonable possibility that a state court would rule against the non-diverse defendant." Poulos v. Naas Foods, Inc. , 959 F.2d 69, 73 (7th Cir. 1992). "If a state court has come to judgment," the federal court must ask whether "there [is] any reasonable possibility that the judgment will be reversed on appeal." Ibid. ; see also Knudson v. Sys. Painters, Inc. , 634 F.3d 968, 980 n.11 (8th Cir. 2011) ; Arthur v. E.I. du Pont , 798 F. Supp. 367, 369 (S.D. W. Va. 1992).3
*297In this case, the state court had granted C.E.N.'s motion for summary judgment. When C.E.N. won summary judgment, it exited the case. See, e.g. , Poulos , 959 F.2d at 70 (describing a party who won summary judgment in state court as "gone from the case"); Higgins v. E.I. DuPont de Nemours & Co. , 863 F.2d 1162, 1166 (4th Cir. 1988) (explaining a "non-diverse party is eliminated from the state action by voluntary or involuntary dismissal" in state court). The federal district court concluded there was no possibility the state court's grant of summary judgment to C.E.N. would be reversed on appeal. The Hoyts had not argued they "could or would pursue an appeal of the granting of C.E.N.'s motion for summary judgment," and the district court concluded the Hoyts "apparently abandoned the right to do so."
The Hoyts' appellate briefs did not challenge the district court's conclusion that they could not appeal the grant of summary judgment to C.E.N. Any argument undermining the district court's conclusion is therefore forfeited. See Cinel v. Connick , 15 F.3d 1338, 1345 (5th Cir. 1994). The Hoyts' counsel did raise the issue at oral argument, but "[a]rguments presented for the first time at oral argument are [forfeited]." Comsat Corp. v. FCC , 250 F.3d 931, 936 n.5 (5th Cir. 2001).
When a state court order creates diversity jurisdiction and that order cannot be reversed on appeal, our precedent treats the voluntary-involuntary rule as inapplicable. See Crockett , 436 F.3d at 533 ("[R]emoval on the basis of an unappealed severance, by a state court, of claims against improperly joined defendants is not subject to the voluntary-involuntary rule."); 14C CHARLES ALAN WRIGHT ET AL., FEDERAL PRACTICE & PROCEDURE § 3723 (Rev. 4th ed. 2019) (explaining Crockett "makes sense because there is no risk of reversal of an unappealed severance order"). In light of the district court's now-unchallengeable ruling that the state court's dismissal of C.E.N. could not be reversed on appeal, we are compelled to conclude C.E.N. was improperly joined.4 As a result, the voluntary-involuntary rule did not bar Lane from removing the case to federal court. Therefore, the district court did not err in denying the Hoyts' second motion to remand.
III.
We now consider whether the district court erred by granting Lane's motion for *298summary judgment. Applying de novo review, DeVoss v. Sw. Airlines Co. , 903 F.3d 487, 490 (5th Cir. 2018), we conclude it did.
A.
The district court first held Lane is statutorily immune from the Hoyts' damages claims. Texas law provides:
A contractor who constructs or repairs a highway, road, or street for the Texas Department of Transportation [ ("TxDOT") ] is not liable to a claimant for personal injury, property damage, or death arising from the performance of the construction or repair if, at the time of the personal injury, property damage, or death, the contractor is in compliance with contract documents material to the condition or defect that was the proximate cause of the personal injury, property damage, or death.
TEX. CIV. PRAC. & REM. CODE § 97.002. According to the district court, Lane performed its work on FM 2264 pursuant to a TxDOT contract and "did everything to follow TxDOT plans."
We disagree. On October 27, 2015-nine weeks before Jeffery's tragic accident-Lane performed remedial "erosion control" work alongside FM 2264. TxDOT determined this remedial work was necessitated by Lane's "negligence." Lane was not properly removing "accumulated silt" from the construction site and instead was "push[ing] off [the silt] along the [right of way]." The Hoyts argue this silt caused the pooling of water on FM 2264 that killed Jeffery. To support that argument, the Hoyts submitted expert testimony that Lane's "[f]ailure to properly install and maintain sediment control measures" in accordance with its TxDOT contract caused water to "flow[ ] onto the roadway."
But that's not all. On January 15, 2016-two weeks after Jeffery's tragic death-TxDOT completed a "Construction Stage Gate Checklist" for the FM 2264 project. It found Lane failed to install stormwater run-off "control measures" in accordance with TxDOT specifications. It further found the "control measures" were not "performing" or "properly maintained."
These record materials-spanning the time period from before Jeffery's accident to after it-raise a material dispute of fact regarding Lane's compliance or lack thereof with its TxDOT contract. Moreover, statutory immunity under § 97.002 is an affirmative defense on which Lane bore the burden of proof. See Brown v. RK Hall Constr., Ltd. , 500 S.W.3d 509, 512 (Tex. App.-Texarkana 2016, pet. denied). And Lane failed to show as a matter of law that its TxDOT-documented negligence before the accident and its TxDOT-documented inspection deficiencies after the accident did not cause Jeffery's death. Therefore, it was error to grant summary judgment on this ground.
B.
The district court next granted summary judgment to Lane on the Hoyts' premises-liability claim. It did so on two grounds. We conclude material disputes of fact preclude both.
First, the district court found the ice patch that killed Jeffery was "a natural formation." It is true that, under Texas law, "naturally occurring ice that accumulates without the assistance or involvement of unnatural contact is not an unreasonably dangerous condition sufficient to support a premises liability claim." Scott & White Mem'l Hosp. v. Fair , 310 S.W.3d 411, 414 (Tex. 2010). Here, however, the Hoyts have pointed to summary judgment evidence that the ice patch on FM 2264 accumulated through unnatural conditions-namely, Lane's construction activities at the crash site. It submitted an *299expert report to that effect. Moreover, the State Trooper who responded to the accident testified the ice patch that killed Jeffery was the only patch of ice along a 20-mile stretch of highway. The second driver (who also slid off the road and landed on top of Jeffery's vehicle) testified to the same effect. This testimony-viewed in the light most favorable to the Hoyts-creates a material dispute of fact regarding whether the ice patch formed naturally.
Second, the district court found Lane "did not have actual knowledge of a dangerous premises condition at the time of [Jeffery's] accident." Under Texas law, actual knowledge can be proved through direct evidence or reasonable inference. For example, in City of San Antonio v. Rodriguez , 931 S.W.2d 535 (Tex. 1996) (per curiam), the plaintiff slipped on a city-owned basketball court. See id. at 536. The city had actual knowledge of leaks in the roof but disputed it had actual knowledge that rain had leaked onto the basketball court when Rodriguez slipped. See id. at 537. The Supreme Court of Texas held that was irrelevant:
[T]here was evidence that the person in charge of the recreation center knew of the leaks in the roof and knew that it had been raining. Depending on the position of the leaks above the floor and the amount of rain, the jury might have inferred that the person in charge knew that there would be water on the floor.
Ibid.
So here. The Hoyts produced evidence tending to show Lane knew the culverts at its construction site could overflow in a heavy rain. Lane knew about the drainage problems associated with its work on the culverts. And Lane knew the site received 4-6 inches of heavy rain, under freezing temperatures, the night before Jeffery died. That distinguishes this case from City of Corsicana v. Stewart , 249 S.W.3d 412 (Tex. 2008) (per curiam), in which the defendant city did not know a road had flooded, in part because a city official was unaware of the rain. See id. at 415-16 & n.3. Therefore, as in Rodriguez , a jury reasonably could infer Lane knew there would be ice on the road. If anything, that inference is easier here because Lane employees repeatedly worried drivers could slip off the road where Jeffery died-with or without ice. As one Lane employee testified, "I just talked to [TxDOT officials] numerous times that I was afraid that our people working there [in the spot where Jeffery died], a car could drive off on top of them."
C.
Finally, the district court granted summary judgment for Lane on the Hoyts' gross negligence claim. The Hoyts argue all the evidence they offered to prove premises liability also suffices to show gross negligence. Having vacated the grant of summary judgment on premises liability, we similarly vacate the grant of summary judgment on gross negligence for reconsideration.5
*300* * *
The district court did not err by denying the Hoyts' motions for remand but did err by granting Lane's motion for summary judgment. The judgment is AFFIRMED in part, VACATED in part, and REMANDED for further proceedings consistent with this opinion.
This Court has "adopt[ed] the term 'improper joinder' as being more consistent with the statutory language than the term 'fraudulent joinder,' " but "there is no substantive difference between the two terms." Smallwood v. Ill. Cent. R.R. Co. , 385 F.3d 568, 571 n.1 (5th Cir. 2004) (en banc). Thus, we use the term "improper joinder" except where quoting other authorities.
The Hoyts argue we should not consider the improper-joinder exception because Lane forfeited the issue by failing to include it in its notice of removal. A notice of removal must "contain[ ] a short and plain statement of the grounds for removal." 28 U.S.C. § 1446(a). Lane's notice of removal identified diversity jurisdiction as the ground for removal and explained that complete diversity existed because the Texas defendants were no longer in the case. Lane first raised improper joinder in its response to the Hoyts' second motion to remand. In this case at least, improper joinder was not a ground for removal. It was a rebuttal to the Hoyts' argument that the voluntary-involuntary rule prevented removal. In other words, it was an exception to an exception to the ground of removal. We do not require a litigant to anticipatorily rebut all potential arguments his adversary may raise. Failing to do so is not a forfeiture. See R.J. Corman Derailment Servs., LLC v. Int'l Union of Operating Eng'rs, Local Union 150 , 335 F.3d 643, 650 (7th Cir. 2003) ("[A] party cannot waive something that it does not know is at issue."); NLRB v. NPC Int'l, Inc. , No. 13-0010, 2017 WL 634713, at *5 (W.D. Tenn. Feb. 16, 2017) ("A moving party does not waive reply arguments merely because it fails to accurately predict the contentions of a responding party.").
Neither the Hoyts nor the dissent point to any cases holding to the contrary. Some courts have held a removing defendant forfeits an improper-joinder argument by failing to include it in the notice of removal when a non-diverse party remains in the case at the time of removal. See, e.g. , Wessel v. Miraglia , No. 4:04-cv-377, 2004 WL 1943776, at *3 (N.D. Tex. Aug. 31, 2004) (treating improper joinder as a ground that should have been included in a notice of removal). But that rule does not apply where, as here, improper joinder is used as an exception to the voluntary-involuntary rule. Thus, we need not decide whether those cases are rightly decided. See Pharos Capital Grp., LLC v. Nutmeg Ins. Co. , 999 F. Supp. 2d 947, 953 (N.D. Tex. 2014) ("[E]ven though it was brought up for the first time in its response to the Motion to Remand, [the removing defendant's] improper joinder argument was not waived.").
Apparently rejecting this framework, the dissent would conduct a "Rule 12(b)(6)-type analysis" and disregard the state court's grant of summary judgment. Post at 302-03 (Haynes, J., dissenting). Under this Court's precedent, a "Rule 12(b)(6)-type analysis" applies when a federal court is considering whether to dismiss a non-diverse defendant as improperly joined. Flagg v. Stryker Corp. , 819 F.3d 132, 136 (5th Cir. 2016) (en banc); Smallwood , 385 F.3d at 572-73. But it does not necessarily apply when a state court has already dismissed the nondiverse defendant. Cf. Crockett , 436 F.3d at 533. It would be passing strange for a federal court asking whether the plaintiff can "establish a cause of action against the nondiverse party in state court," Flagg , 819 F.3d at 136 (quotation omitted), to ignore that "a state court has come to judgment" on that very question, Poulos , 959 F.2d at 73 ; see also Knudson , 634 F.3d at 980 n.11. "We are always chary to create a circuit split," Gahagan v. U.S. Citizenship & Immigration Servs. , 911 F.3d 298, 304 (5th Cir. 2018) (quotation omitted), and the dissent provides no compelling reason to do so here. The bottom line is this: for the reasons we have explained, the Hoyts' case against C.E.N. was not winnable because they could not appeal the state court's dismissal.
The dissent says the Hoyts' forfeiture is irrelevant because the issue is "jurisdictional." Post at 300-01 n.1 (Haynes, J., dissenting). That conflates the rules of subject-matter jurisdiction (which cannot be forfeited) with the rules of removal procedure (which can be forfeited). Only Congress can make rules of unforfeitable subject-matter jurisdiction. See, e.g. , Hamer v. Neighborhood Hous. Servs. of Chi. , --- U.S. ----, 138 S. Ct. 13, 17, 199 L.Ed.2d 249 (2017). By contrast, the judge-made voluntary-involuntary rule is a forfeitable rule of removal procedure. See Crockett , 436 F.3d at 533 (describing the voluntary-involuntary rule as a matter of "removal jurisdiction"); Baris v. Sulpicio Lines, Inc. , 932 F.2d 1540, 1543 (5th Cir. 1991) (explaining "improper removal (i.e., lack of removal jurisdiction) ... is waivable"); see also Higgins , 863 F.2d at 1166 (holding a potential violation of the voluntary-involuntary rule did not deprive a federal district court of jurisdiction); Jones v. CMM of Ind., LLC , No. 05-cv-4039, 2006 WL 120085, at *2 n.2 (E.D. La. Jan. 17, 2006) (holding "the voluntary/involuntary rule is one of removal procedure rather than subject matter jurisdiction"). In all events, the Hoyts cannot bring C.E.N. back into the case for an independent reason-namely, their claims against C.E.N. are meritless for all the reasons the district court already gave. See Hoyt v. Lane , Order on Plaintiffs' Second Motion to Remand, No. 17-cv-780, Doc. 32, at 4-5 (Dec. 4, 2017). Their "inability ... to establish a cause of action against the non-diverse party in state court" demonstrates improper joinder by itself. Crockett , 436 F.3d at 532 (quotation omitted).
That reconsideration need not include the Hoyts' alternative argument that Lane's concessions regarding material disputes of fact in state court preclude summary judgment in federal court. Neither of the doctrines the Hoyts invoke-judicial admissions and evidentiary admissions-applies to concessions of law. "[W]hether a genuine dispute concerning a material fact exists is itself a question of law that must be decided by the court. It does not depend upon what either or both of the parties may have thought about the matter." 10A Charles Alan Wright et al., Federal Practice & Procedure § 2720 (4th ed. 2019) ; see also John v. Louisiana , 757 F.2d 698, 712 (5th Cir. 1985). "[A] statement of legal opinion ... is not a judicial admission." McCaskill v. SCI Mgmt. Corp. , 298 F.3d 677, 682 (7th Cir. 2002) ; see also Anderson v. Comm'r of Internal Revenue , 698 F.3d 160, 167 (3d Cir. 2012) ; Roger Miller Music, Inc. v. Sony/ATV Publ'g, LLC , 477 F.3d 383, 394 (6th Cir. 2007) ; N.Y. State Nat'l Org. for Women v. Terry , 159 F.3d 86, 97 n.7 (2d Cir. 1998). "[A]nd legal conclusions may not be used as evidentiary admissions." W.V. Realty, Inc. v. N. Ins. Co. of N.Y. , 334 F.3d 306, 316 (3d Cir. 2003) ; see also Incase Inc. v. Timex Corp. , 488 F.3d 46, 56 (1st Cir. 2007). Therefore, Lane's state court concessions do not help the Hoyts here.