Harnessing nuclear energy is a delicate business. So is the statute before us. Originally passed in the 1950s in an era captivated by the promise of nuclear power and amended in the 1980s in the aftermath of the Three' Mile Island meltdown when prevailing public sentiment was perhaps less sanguine, the Price-Anderson Act seeks both to promote the private nuclear energy industry and, simultaneously, to ensure relief for those injured by it. In this appeal, we consider how far Congress went in reshaping state tort claims involving what the Act delicately refers to as nuclear “incidents” and “occurrences”— and what our own prior encounter with this case has to say on the subject.
*1090I
The beginnings of our dispute trace back generations. During the Cold War, the Rocky Flats plant served as home to a nuclear weapons production facility. Located just sixteen miles from -downtown Denver, the plant was operated first by Dow, then by Rockwell, under contracts with the federal government. But everything ground to a halt in 1989. That’s when FBI agents raided the plant and unearthed evidence of environmental crimes. It turns out plant workers had mishandled radioactive waste for years. Some had been poured into the ground and leached into nearby bodies of water. Some had been released into the air and filtered its way into the soil throughout the area. As news of all this emerged, the plant’s neighbors saw their property values plummet. And soon enough they followed the government’s criminal action with a civil suit of their own, citing both the federal Price-Anderson Act and state nuisance law as grounds for relief.
It took a titanic fifteen years for the case to reach a jury. No doubt a testament to contemporary civil litigation practices that ensure before any trial is held every stone will be overturned in discovery — even if it means forcing everyone to endure the sort of staggering delay and (no doubt) equally staggering expense the parties endured here. Somehow, though, this case managed to survive the usually lethal gauntlet of pretrial proceedings and stagger its way to trial. There the jury found for the plaintiffs and the district court approved roughly $177 million in compensatory damages and $200 million in punitive damages — as well as $549 million in prejudgment interest, thanks again to all the pretrial delay.
That, though, was hardly the' end of it. Next the case found its way to appeal — for the first time. On appeal, the defendants argued that the district court had failed to instruct the jury properly about the terms of the Price-Anderson Act. Under the Act, any lawsuit asserting liability for a “nuclear incident” is automatically considered a federal action that can be brought in (or removed to) federal court. See 42 U.S.C. §§ 2014(w), 2014(hh), 2210(n)(2). And if that assertiqn is proven at trial — if the jury finds that a nuclear incident actually occurred and harmed the plaintiffs — a number of special rules kick in, including rules limiting the liability of certain defendants and requiring the government to pay any damages not covered by insurance. See id. § 2210(c)-(e). Unsurprisingly given these generous financial protections, defendants often have as much incentive as plaintiffs to accept that any harm they caused stemmed from a nuclear incident.
But in this particular case Dow and Rockwell made a curious tactical decision. They argued that the district court’s jury instructions about what constitutes a nuclear incident were too permissive. To prove that a nuclear incident has damaged real property the Act requires a plaintiff to show the “loss of or damage to property, or loss of use of property.” Id. § 2014(q). According to the defendants, this meant the district court should have required the plaintiffs to prove at trial physical damage to their property or the loss of a specific, particularized use of their property — not mere contamination by radioactive materials or reduced property values. Of course, in the long run an argument along these lines promised to restrict the Act’s application, including the benefits it affords defendants. But in the short run it offered a way to overturn the district court’s judgment in this case. And Dow and Rockwell leapt at the chance.
The defendants’ tactical decision seemed to pay off. This court agreed that the district court’s jury instructions about what does and doesn’t qualify as a nuclear *1091incident were too permissive. On this basis, it vacated the district court’s judgment and remanded the case for further proceedings in light of the Act’s correct construction. Cook v. Rockwell Int’l Corp., 618 F.3d 1127, 1138-42, 1153 (10th Cir.2010).
But that’s when things took an interesting turn. Trying their hand at a little judicial jiu-jitsu, the plaintiffs sought to turn the defendants’ victory against them. Back before the district court on remand the plaintiffs disclaimed any effort to prove a nuclear incident for purposes of the Price-Anderson Act. Forget the Act and the benefits it provides to both sides, they said, we renounce them. Accepting the premise that they couldn’t prove a nuclear incident — at least as the term was interpreted by this court — the plaintiffs argued this meant only that the Act’s liability limiting and indemnification protections fall away, leaving background state tort law to operate normally.1 What’s more, the plaintiffs submitted, everything needed for a judgment on a state law nuisance claim already existed. The operative complaint expressly sought relief under Colorado nuisance law. At trial, the district court instructed the jury on Colorado nui-sanee law and the jury returned a state law nuisance verdict in accordance with those instructions. In the first appeal, this court held that “[t]he jury was properly instructed on the elements of a nuisance claim.” Id. at 1145. And no one has ever challenged the sufficiency of the evidence in the record. The state law nuisance verdict, moreover, was untainted by any error identified in the first appeal: this court reversed only because of an instructional error concerning' what’s needed to prove a nuclear incident under the Act, not what’s required to prove a nuisance under Colorado law. So, the plaintiffs argued, all the ingredients required for a state law nuisance judgment were present and accounted for and a judgment on that claim should issue forthwith. Perhaps the defendants’ push in the first appeal for a narrow definition, of what qualifies as a nuclear incident won them the battle, but it lost them the war — failing to eliminate the plaintiffs’ state law claim and serving only to narrow now and in the future both sides’ ability to secure the benefits of the Price-Anderson Act.2
This put the defendants in a pickle. Having prevailed on the argument that this case doesn’t involve a nuclear incident *1092sufficient to implicate the federal statutory regime, now they had to conjure some reason why entering a judgment on the existing state law nuisance verdict would be legally impermissible. Ultimately the defendants settled on two arguments. First, they contended, the Price-Anderson Act prevents it. While everyone agrees that the Act provides a federal forum when a nuclear incident is asserted, and that it affords certain liability protections and guarantees if a nuclear incident is proven, Dow and Rockwell now suggested the Act does something more and entirely different. They argued to the district court that the Act also preempts and precludes any state law recovery where (as here) a nuclear incident is asserted but ultimately unproven. In this way, the defendants suggested, the Act embodies a sort of go-big- or-go-home rule of liability. If you allege and successfully prove a full-blown nuclear incident your recovery may be assured by the full faith and credit of the federal government. But if you allege and then fail to prove a nuclear incident you are barred from recovery of any kind — even if you can establish a qualifying state law nuisance. Second, Dow and Rockwell argued that this court’s mandate in the first appeal independently barred the plaintiffs from securing relief on their existing state law nuisance verdict. So even if the plaintiffs’ state law claim isn’t preempted any recovery is just as effectively foreclosed.
Ultimately, the district court sided with the defendants on both points and so here we are again. This time on the plaintiffs’ appeal asking us to reverse the district court’s preemption ruling and its holding about the scope of this court’s mandate.
II
We begin with the preemption question. Preemption can come about in various ways and it’s no simple thing or our ambition to attempt a complete taxonomy. Still, it’s fair to say the Supreme Court has distinguished between “express” preemption (where Congress explicitly indicates its intent to supplant state law) and “implied” preemption (where some other aspect of a statute is .said to suggest such an intent). See, e.g., Altria Grp., Inc. v. Good, 555 U.S. 70, 76-77, 129 S.Ct. 538, 172 L.Ed.2d 398 (2008). And it’s fair to say the Court has distinguished between “field” preemption (where Congress leaves “no room” for state regulation in an entire area) and “conflict” preemption (where Congress has expressed a more modest wish to displace individual state laws standing in the way of federal law). Arizona v. United States, — U.S. —, 132 S.Ct. 2492, 2500-01, 183 L.Ed.2d 351 (2012).
Of course, these labels “are anything but analytically air-tight.” Laurence H. Tribe, American Constitutional Law § 6-28, at 1177 (3d ed.2000). But they do shed some light on what arguments are — and aren’t— before us. Examining the defendants’ submissions, it’s clear that Dow and Rockwell fail to invoke implied preemption doctrine and even appear to disclaim reliance on it. See Appellees’ Br. 35. Neither do the companies depend on more modest conflict preemption principles — and potential preemption defenses, like most other affirmative defenses, are forfeited if not made. See, e.g., Mauldin v. Worldcom, Inc., 263 F.3d 1205, 1211 (10th Cir.2001). So the only preemption argument at issue before us is the suggestion that the Price-Anderson Act expressly preempts and precludes relief for the field of claims that assert but fail to prove a nuclear incident. And with our hands now around the parameters of the argument before us, we can see it quickly encounters two separate and independently dispositive problems. See Woods v. Interstate Realty Co., 337 U.S. 535, 537, 69 S.Ct. 1235, 93 L.Ed. 1524 (1949).
*1093A
The first problem is a procedural one: the defendants forfeited any contention along these lines in the first appeal. At that point the plaintiffs had a judgment in their favor under both state and federal law. The defendants argued that the federal component of the judgment was infected by instructional error and required a remand, but they never suggested a freestanding state tort judgment would be preempted and precluded. In fact, this court explained that though the defendants had “allude[d] to field preemption in their brief’ in the first appeal, they “never developed] the issue.” Cook, 618 F.3d at 1143 n. 16. And just as it’s important to raise a preemption defense or risk losing it, it’s important to raise it in a timely manner. This court has long explained that non-jurisdictional arguments a party forfeits on appeal “may not be asserted ... on remand.” Dow Chem. Corp. v. Weevil-Cide Co., 897 F.2d 481, 486 n. 4 (10th Cir.1990). So it appears the defendants had no business attempting a field preemption affirmative defense following the first appeal, and the district court had no business adopting it. After all, at some point it becomes too late to try out promising new theories, gangly litigation must be shaped and defined, and finality has to become more than a faint hope. Maybe especially when it comes to a new defense raised on remand some twenty-five years after the case began.
Dow and Rockwell don’t dispute these legal principles, only their application to this case. Whatever the first panel may have said, they insist they have always presented a field preemption defense to any freestanding state law nuisance claim, including in the first appeal. The defendants note that the district court on remand agreed with their self-assessment on this score. But under law of the case doctrine what governs is the first panel’s holding that the defendants failed to develop the argument — not the defendants’ or the district court’s insistence otherwise on remand — for in our legal order the parties and trial courts aren’t usually free to revisit issues an appellate court has resolved. Neither do the defendants attempt to invoke any recognized exception to the doctrine that might allow us to upset this, the normal order of things. See United States v. Alvarez, 142 F.3d 1243, 1247 (10th Cir.1998).
Instead, the defendants reply that the first panel spoke at most to an entirely different line of argument than the one they now seek to raise. According to Dow and Rockwell, the panel’s holding that they had foregone any field preemption argument wasn’t directed to the field preemption question they currently wish to pursue (does the Act preclude any state forms of relief when a nuclear incident is asserted but unproven?), but to a separate preemption question (when must state tort law standards of care give way in a claim because they conflict with particular terms found in the Price-Anderson Act?).
We admit this may not be a frivolous reading of the prior panel’s opinion, but it is surely an odd one. The term “field preemption,” after all, doesn’t refer to situations where some specific conflict exists between federal and state rules of law but to situations where the federal government has so fully occupied an entire field that no room remains for the operation of state law at all. Conflict on a grand scale, not a discrete one. See supra at 1092-93. Yet under the defendants’ reading of the first panel’s decision we must assume that the panel was confused about this much and found forfeited a field preemption argument about what is at most a conflict preemption question. Unsurprisingly, we are unwilling to give such an uncharitable gloss to our colleagues’ handiwork and *1094reaffirm the first panel’s ruling that Dow and Rockwell forfeited any field preemption argument long ago.
B
Even if we were to overlook this procedural problem a substantive one would soon emerge for we see no field preemption in the Act of the sort the defendants describe. When it comes to the preemptive force of the Act the parties present us with two very different narratives: one entirely familiar, the other more than a little incongruous. On the plaintiffs’ reading, the Act embodies an arrangement much like that found, in the Class Action Fairness Act and similar statutes, one in which Congress hasn’t preempted an entire field but provided a federal forum and certain specific rules for larger cases while allowing smaller cases more or less to go their own way. See 28 U.S.C. § 1332(d); Estate of Pew v. Cardarelli, 527 F.3d 25, 26 (2d Cir.2008). On the defendants’ reading, meanwhile, the Act guarantees recovery for plaintiffs who plead and prove a nuclear incident even as it preempts and precludes any recovery for plaintiffs who plead a nuclear incident but ultimately prove only a lesser state law nuisance. So on the defendants’ view an affirmative defense (field preemption) is, literally, complete and effective upon a plaintiffs mistaken assertion of fact (the existence of a nuclear incident). Maybe this regime the defendants imagine isn’t metaphysically impossible but stating it in concise terms does make you wonder. A wonder that grows even larger when you consider the defendants’ suggestion that their affirmative defense would serve to preclude small claims even as the statute it stems from admittedly guarantees recovery for larger ones.
When confronted with competing preemption narratives like these the Supreme Court has instructed us to “start with the assumption that the historic police powers of the States were not to be superseded ... unless that was the clear and manifest purpose of Congress.” Rice v. Santa Fe Elevator Corp., 331 U.S. 218, 230, 67 S.Ct. 1146, 91 L.Ed. 1447 (1947). So to the extent Congress’s statutory direction is susceptible to more than one reading, we have the “duty to accept the reading that disfavors pre-emption.” Bates v. Dow Agrosciences LLC, 544 U.S. 431, 449, 125 S.Ct. 1788, 161 L.Ed.2d 687 (2005). A duty that is only “heightened” where (as here) the area of law in question is one of traditional state regulation like public health and. safety. Riegel v. Medtronic, Inc., 552 U.S. 312, 334, 128 S.Ct. 999, 169 L.Ed.2d 892 (2008). These presumptions seek to ensure “that the federal-state balance will not be disturbed unintentionally by Congress or unnecessarily by the courts.” Id. (quoting Jones v. Rath Packing Co., 430 U.S. 519, 525, 97 S.Ct. 1305, 51 L.Ed.2d 604 (1977)) (internal quotation marks omitted). And applying the traditional tools of statutory interpretation to the Act before us, it quickly becomes clear that nothing in its language, structure, or history favors the defendants’ curious statutory construction over the plaintiffs’ prosaic one — let alone favors it so clearly that we might overcome the presumption against preemption.
Start with the text. The Priee-Anderson Act applies to “any suit asserting public liability.” 42 U.S.C. § 2014(hh). It defines “public liability” to mean “any legal liability arising out of or resulting from a nuclear incident.” Id. § 2014(w). A “nuclear incident,” meanwhile, refers to “any occurrence” that causes “bodily injury, sickness, disease, or death, or loss of or damage to property, or loss of use of property, arising out of or resulting from the radioactive, toxic, explosive, or other hazardous properties” of nuclear materials. Id. § 2014(q). Together, then, “a ‘public *1095liability action’ is a suit in which a party asserts that another party bears any legal liability arising out of an incident in which the hazardous properties of radioactive material caused bodily injury, sickness, or property damage.” Cotroneo v. Shaw Env’t & Infrastructure, Inc., 639 F.3d 186, 194 (5th Cir.2011). And, as we’ve seen, special rules apply when a party pursues a public liability action. For starters, the suit is “deemed to be an action arising under” federal law: not only may plaintiffs file their complaints in federal court, but defendants may remove any qualifying suit from state court, too. 42 U.S.C. § 2014(hh); see id. § 2210(n)(2). If a plaintiff goes beyond merely asserting the existence of a nuclear incident- — -actually convincing the factfinder such an incident took place — the Act proceeds to limit the amount of liability certain defendants may face and it obligates the government to underwrite certain of these losses. Id. § 2210(c)-(e). All the while, though, the Act stipulates that state law provides the “substantive rules for decision” in the action except to the extent state law proves “inconsistent” with the terms of § 2210. Id. § 2014(hh).
Where does any of this language — expressly — preempt and preclude all state law tort recoveries for plaintiffs who plead but do not prove nuclear incidents? We just don’t see it. Congress knows well how to preempt a field expressly when it wishes. In the Occupational Safety and Health Act, for example, it explicitly required states wishing to regulate in the field to submit their plans for approval by the Secretary of Labor. See Gade v. Nat’l Solid Wastes Mgmt. Ass’n, 505 U.S. 88, 111-13, 112 S.Ct. 2374, 120 L.Ed.2d 73 (1992) (Kennedy, J., concurring in part and concurring in the judgment). There’s just nothing like that in the statutory text before us.
The defendants insist § 2014(hh) does the job. But, again, that section merely affords 'a federal forum when a nuclear incident is “assert[edj” and provides a modest form of conflict preemption once the • case is underway: normal state law principles continue to govern unless they conflict with the rules found in § 2210. So, for example, once a nuclear incident is proven § 2210(e)’s limitations on “aggregate public liability for a single nuclear incident” apply, notwithstanding the availability of greater damages under state law. Nothing in this language speaks to what happens when a nuclear incident is alleged but unproven. And certainly nothing in it dictates that injured parties in such circumstances are forbidden from seeking or securing traditional state law remedies.
Surrounding textual features confirm the point. Consider § 2014(q), which defines the term “nuclear incident” to mean an “occurrence” that causes “bodily injury, sickness, disease, or death, or loss of or damage to property, or loss of use of property.” From this language, it’s clear Congress anticipated the possibility of lesser nuclear “occurrences” that fail to rise to the level of nuclear “incidents” — it even gave them a name. And in that light, the absence of any accompanying statutory bar to state tort recovery in cases involving these lesser “occurrences” takes on an even more deliberate hue. Consider as well § 2014(w), which defines “public liability” to mean “any legal liability arising out of or resulting from a nuclear incident.” 42 U.S.C. § 2014(w). From this, it appears that once injuries sufficient to trigger a nuclear incident finding are proven the Act contemplates recovery (and liability limitations and indemnification) for “any ” injuries flowing from that incident, even those that aren’t themselves sufficient to trigger a nuclear incident finding. And it’s hard to conjure a reason why Congress would allow plaintiffs to recover *1096for a full panoply of injuries in the event of a large nuclear incident but insist they get nothing for a lesser nuclear occurrence. Certainly the defendants don’t offer this court any rationale that might explain such a rule.
Looking beyond textual clues to take in the larger statutory structure does nothing to alter our impression. Not only can federal claims for larger nuclear incidents subject to the Act’s limitations and benefits coexist with state law claims for lesser nuclear occurrences, they can do so quite sensibly. Larger occurrences that qualify as nuclear incidents can threaten to bankrupt nuclear power providers and leave victims un-(or under-) compensated. In these cases, it’s understandable why Congress might intercede to provide liability caps and indemnification. Meanwhile, smaller occurrences are less likely to raise the same concerns, so it’s equally understandable why Congress might not prevent state law from running its course with respect to them. Our case illustrates the point. At trial, Dow and Rockwell’s liability for compensatory damages totaled roughly $177 million. The Act, meanwhile, currently caps federal contractors’ liability for nuclear incidents at approximately $12.7 billion. See id. § 2210(d)(2), (e)(1)(B); Adjustment of Indemnification for Inflation, 78 Fed.Reg. 56,868 (Sept. 16, 2013). Indeed, even if we count the $200 million in punitive damages and $549 million in prejudgment interest — for a total of $926 million — we still come only 7.29% of the way to the $12.7 billion cap. The claims here thus simply do not appear to be of the sort that implicate the Act and its textually manifest concerns related to liability limitation and indemnification.
A study of the Act’s history yields still more evidence pointing in the same direction. There’s much in that history suggesting Congress passed the Act to improve the manageability of complex litigation, to ensure that liabilities arising from large nuclear incidents don’t shutter the nuclear industry, and to guarantee compensation for victims who otherwise might be left trying to squeeze damages out of firms bankrupted by enormous awards. See, e.g., Duke Power Co. v. Carolina Envtl. Study Grp., Inc., 438 U.S. 59, 64, 98 S.Ct. 2620, 57 L.Ed.2d 595 (1978) (noting that “the risk of potentially vast liability in the event of a nuclear accident of a sizable magnitude ” motivated Congress to pass the Act (emphasis added)); id. at 83, 98 S.Ct. 2620 (“As we read the Act and its legislative history, it is clear that Congress’ purpose was ... to stimulate the private development of electric energy by nuclear power while simultaneously providing the public compensation in the event of a catastrophic nuclear incident.” (emphasis added)). Meanwhile, little in the Act’s history suggests an intent to preclude recovery or inhibit the operation of state tort law in cases involving lesser nuclear occurrences that don’t give rise to the sorts of injuries and damages involved in more serious nuclear incidents. Indeed, the evidence suggests that Congress sought to “minimize] interference with State law” so that “the only interference with State law is ... in the exceedingly remote contingency of a nuclear incident giving rise to damages in excess of the amount of financial responsibility required together with the amount of the governmental indemnity.” S.Rep. No. 89-1605, at 6 (1966); see also H.R.Rep. No. 100-104, pt. 1, at 20 (1987).
The backdrop to the Act’s 1988 amendments is consistent with this understanding, too. It appears Congress passed those amendments, which included § 2014(hh), in an effort to smooth and speed the recovery process for victims after witnessing the aftermath of the Three Mile Island incident, which resulted in a fair amount of litigation chaos: “over 150 *1097separate cases against TMI defendants, with over 3,000 claimants, in various state and Federal courts.” S.Rep. No. 100-218, at 13 (1987), 1988 U.S.C.C.A.N. 1476, 1488. Yet, as one of our colleagues has pointed out, Dow and Rockwell’s reading of the law (no recovery absent a full-blown nuclear incident) would have the surprising effect of barring recovery “in the event of a future accident exactly like Three Mile Island,” because “Three Mile Island does not appear to have caused” the sort of grave injuries required to establish a nuclear incident under § 2014(q). Cotroneo, 639 F.3d at 206 (Dennis, J., concurring in part and dissenting in part).
Even if nothing in the language, structure, or history of the Price-Anderson Act allows us to overcome the presumption against preemption and adopt the defendants’ unlikely reading of the Act, Dow and' Rockwell argue Supreme Court precedent requires us to reach that result all the same. Here they rest heavily on El Paso Natural Gas Co. v. Neztsosie, 526 U.S. 473, 119 S.Ct. 1430, 143 L.Ed.2d 635 (1999), a decision in which the Supreme Court referred to § 2014(hh) as an “unusual preemption provision” and stated that the Act’s “structure, in which a public liability action becomes a federal action, but one decided under substantive state-law rules of decision that do not conflict with the Price-Anderson Act, resembles what we have spoken of as ‘complete pre-emption doctrine.’ ” Id. at 484 & n. 6, 119 S.Ct. 1430 (citation omitted) (quoting Caterpillar Inc. v. Williams, 482 U.S. 386, 393, 107 S.Ct. 2425, 96 L.Ed.2d 318 (1987)).
We don’t see anything in these remarks at odds with what we’ve said. As Neztsosie indicated, the Act’s operation does indeed “resemble” what’s sometimes called complete preemption. A subspecies of field preemption, complete preemption arises when Congress affords defendants not only an affirmative defense against state law claims, but also the right to remove the dispute to federal court — ensuring that the preemption question itself is decided in a federal (rather than a state) forum. See Beneficial Nat’l Bank v. Anderson, 539 U.S. 1, 6-8, 123 S.Ct. 2058, 156 L.Ed.2d 1 (2003); Devon Energy Prod. Co. v. Mosaic Potash Carlsbad, Inc., 693 F.3d 1195, 1203-04 & n. 4 (10th Cir.2012). And we do have something like that here: the Act provides a federal forum for cases asserting liability arising out of a nuclear incident. At the same time, though, Neztsosie was right to suggest that the Act is “unusual” compared to true complete preemption statutes. For while the Act provides a federal forum it also does much to preserve state rules of decision — quite unlike in true complete preemption statutes where federal law provides the “exclusive cause of action.” Anderson, 539 U.S. at 8, 123 S.Ct. 2058. Confirming the point that ours is not a true complete preemption statute, a few years after Neztsosie the Supreme Court itself noted that it’s so far encountered only three such statutes— and acknowledged the Price-Anderson Act is not among them. See id. at 8-9, 123 S.Ct. 2058.
Without help from the main authority on which they stake their claim in this appeal, the defendants seek support from other Supreme Court decisions suggesting that “the federal government has occupied the entire field of nuclear safety concerns.” Appellees’ Br. 27 (citing PG & E Co. v. State Energy Res. Conservation & Dev. Comm’n, 461 U.S. 190, 212, 103 S.Ct. 1713, 75 L.Ed.2d 752 (1983); English v. Gen. Elec. Co., 496 U.S. 72, 82, 110 S.Ct. 2270, 110 L.Ed.2d 65 (1990); Silkwood v. Kerr-McGee Corp., 464 U.S. 238, 249, 104 S.Ct. 615, 78 L.Ed.2d 443 (1984)). But the panel in the first appeal already explained why and how the defendants overread these decisions and we see no reason or authority that might allow us to deviate from its *1098explanation. See Cook, 618 F.3d at 1143-44 & n. 17. As the first panel explained, these cases recognize that Congress has authorized the federal government alone to promulgate before-the-fact nuclear safety-regulations but — at the same time — has done little to forbid states from indirectly regulating nuclear safety through the operation of traditional after-the-fact tort law remedies. So, for example, in Silkwood the Supreme Court explained that “Congress’ decision to prohibit the states from regulating the safety aspects of nuclear development” did nothing to undermine the “ample evidence that Congress had no intention of forbidding the states from providing [traditional tort] remedies.” 464 U.S. at 250-51, 104 S.Ct. 615. Indeed, the Court expressly acknowledged that “the discussion preceding [the Price-Anderson Act’s] enactment and subsequent amendment indicates that Congress assumed that persons injured by nuclear accidents were free to utilize existing state tort law remedies.” Id. at 251-52, 104 S.Ct. 615 (footnote omitted). Neither is this arrangement remotely unusual. Often Congress entrusts before-the-fact regulation to a federal agency while leaving at least some room for after-the-fact state law tort suits. It has done so in the field of motor vehicle safety. See, e.g., Geier v. Am. Honda Motor Co., 529 U.S. 861, 867-68, 120 S.Ct. 1913, 146 L.Ed.2d 914 (2000). It has done so in the field of medical devices. See, e.g., Caplinger v. Medtronic, Inc., 784 F.3d 1335, 1337-38 (10th Cir.2015). And all the statutory evidence before us suggests it has done the same thing here.
Lacking any Supreme Court precedent to support their atextual result, Dow and Rockwell are left to lean on cases from other circuits. But like Neztsosie, most of these decisions just don’t address the question before us. One case on which the defendants place great emphasis simply says that “[t]he PAA is the exclusive means of compensating victims for any and all claims arising out of nuclear incidents.” In re Hanford Nuclear Reservation Litig., 534 F.3d 986, 1009 (9th Cir.2007). But precisely no one disputes this beside-the-point point. The issue before us isn’t what happens in the event of a nuclear incident, but (again) what happens in the face of a lesser occurrence. A pair of later Ninth Circuit cases come closer to the mark, holding that “any suit seeking compensation for a nuclear incident is preempted by the Act.” Dumontier v. Schlumberger Tech. Corp., 543 F.3d 567, 571 (9th Cir.2008) (second emphasis added); see also Golden v. CH2M Hill Hanford Grp., Inc., 528 F.3d 681, 683-84 (9th Cir.2008) (same). But by way of support for this notion they cite only Hanford’s holding that the Act is the exclusive means of compensating victims of nuclear incidents, offering nothing to explain how or why the Act might preclude relief in cases involving lesser occurrences.
Perhaps the only case the defendants cite that much helps them is Cotroneo, a split decision from the Fifth Circuit. There the majority did find for the defendants but in doing so failed to identify any provision of the Act that expressly preempts and precludes state law claims in the absence of a nuclear incident. Instead, the court reasoned more generally that to allow parties to recover under state law for lesser occurrences would “circumvent the entire scheme governing public liability actions.” 639 F.3d at 197. Of course, this seems a good bit like an implied preemption argument — a suggestion that state suits offend some underlying statutory policy, not any express statutory language— and thus an -argument Dow and Rockwell appear to have disclaimed in this appeal. See supra at 1092-93. But even on its own terms, we have as much difficulty with this argument as Judge Dennis did in dissent, for we fail to discern how our reading of the Act “circumvents” anything. See Co*1099troneo, 639 F.3d at 200-02 (Dennis, J., concurring in part and dissenting in part). As we’ve seen, there’s nothing inconsistent about a statutory scheme that provides federal jurisdiction over certain claims to ensure their streamlined processing — and that includes special rules like liability caps for a subset of those claims involving nuclear incidents — while permitting claims involving lesser occurrences to proceed to decision under preexisting state law principles. If anything and again, it’s the alternative interpretation of the statute — as foreclosing small claims while guaranteeing larger ones- — that appears a good deal less likely in light of the Act’s language, structure, and history.3
HI
Unpersuaded that the Act preempts and precludes a state law nuisance claim when a nuclear incident is asserted but unproven, we arrive at the second question presented in this appeal: does this court’s mandate in the first appeal require much the same result anyway? On remand, the plaintiffs acknowledged that this court identified errors of federal law in the district court’s jury instructions and a judgment under the Price-Anderson Act was no longer possible. But, they argued, this court identified no error associated with their state law nuisance verdict and so nothing prevented the district court from entering a new judgment on their state law nuisance verdict standing alone. Dow and Rockwell replied that, even putting aside their inventive preemption argument, this court’s mandate in the first appeal precluded the plaintiffs’ proposed course. And, again, the district court agreed with them. But how might this be? As we’ve seen, the defendants don’t suggest that the first panel considered the preemption argument they now offer, adopted it, and ruled the plaintiffs’ state law claim precluded on this basis: much to the contrary, the defendants insist that this court didn’t address their field preemption argument in the first appeal. See supra at 1093. So what exactly is it in the first panel’s decision that prohibited the district court on remand from entering judgment on the plaintiffs’ state law nuisance claim? Dow and Rockwell offer two theories.
A
They begin by suggesting that the first panel “explicitly disapproved of the nuisance theory presented at trial under Colorado law.” Appellees’ Br. 44. Here, then, it turns out we confront a comparatively modest argument about Colorado state nuisance law and this court’s treatment of it in the first appeal. In the defendants’ view, the plaintiffs are not entitled to a judgment on the existing nuisance verdict because this court found the nuisance theory presented at trial legally erroneous under Colorado law. As modest as it is, though, the argument is no more successful for it. Far from “disapproving” the plaintiffs’ state law nuisance theory, the panel in the first appeal held that “[t]he jury was properly instructed on the elements of a nuisance claim.” Cook, 618 *1100F.3d at 1145. Just the opposite of what the defendants contend.
Neither does the defendants’ argument get any better when you try to untangle its many and stringy particulars. When it came to state nuisance law in the first appeal, the defendants spent most of their time attacking a district court pretrial order the parties refer to as Cook IX. See Appellants’ Opening Br. at 4861, Cook, 618 F.3d 1127 (Nos. 08-1224, 08-1226, OS-1239) (citing Cook v. Rockwell Int’l Corp. (Cook IX), 273 F.Supp.2d 1175, 1201-08 (D.Colo.2003)). In that pretrial order the district court suggested that a plaintiffs individualized anxiety or fear might be enough on its own to give rise to a nuisance claim in Colorado. And this court in the first appeal did indicate that the district court’s understanding on this point was in error. Cook, 618 F.3d at 1145-46. But by the time of the first appeal, nothing turned on Cook IX’s pretrial discussion. By then a trial had taken place and the district court had issued jury instructions on the elements of Colorado nuisance law. And it’s the law as instructed to the jury, not the law as discussed in tentative pretrial opinions, that matters. See Lederman v. Frontier Fire Prot., Inc., 685 F.3d 1151, 1155 (10th Cir.2012) (explaining that after trial the relevant inquiry is “whether the jury was misled in any way and whether it had an understanding of'the issues and its duty to decide those issues” (brackets omitted)).
When it comes to the consequential jury instructions, moreover, not only did this court expressly hold them proper, it did so for good reason. The defendants in the first appeal complained that language in Instructions 3.6 and 3.7 might suggest that “some increased health risk” was enough to sustain a state nuisance claim. But courts review jury instructions as a whole, not in “artificial isolation.” Cupp v. Naughten, 414 U.S. 141, 147, 94 S.Ct. 396, 38 L.Ed.2d 368 (1973). And reading the instructions as a whole, it’s impossible to understand them as authorizing a judgment for the plaintiffs based solely on “some increased health risk.” The district court did tell the jury that increased health risk stemming from the defendants’ damage to the plaintiffs’ property could constitute a form of “interfere[nce]” with the plaintiffs’ “use and enjoyment” of their property for purposes of state nuisance law. But Instruction 3.6 went on to say that the plaintiffs also had to prove that the degree of the defendants’ interference was “both ‘unreasonable’ and ‘substantial.’ ” Instruction 3.9 explained that an interference is “substantial” only if a “normal person in the community would find it offensive, annoying or inconvenient.” Instruction 3.10 added that to find an interference “unreasonable” the plaintiffs had to prove that “the gravity of the harm outweigh[ed] the utility of the conduct that caused it.” And Instructions 3.11 and 3.12 set forth a series of factors for the jury to balance when conducting this analysis. Before finding liability, then, the jury had to weigh the overall social utility of the defendants’ actions from an “objective perspective” and couldn’t rely solely on individual plaintiffs’ subjective fears and anxieties or the possibility of “some” increased health risk.4 Given this, it is altogether unsurprising that the first panel held “[t]he jury was properly instructed on the elements of a nuisance claim as well as the definitions of ‘substantial’ and ‘unreasonable.’ ” Cook, 618 F.3d at 1145.
*1101At this point you might wonder why the panel in the first appeal bothered to discuss and disapprove of the incorrect understanding of state nuisance law found in the Cook IX pretrial order when the panel expressly found the different formulation of state nuisance law embodied in the operative jury instructions entirely acceptable. The answer is both simple and sensible: even if the first panel could’ve avoided some of the issues briefed before it, it judged it “proper to nonetheless decide questions of law raised in this appeal that are certain to arise again ... on remand.” Id. at 1142 n. 15. The panel was well aware that the plaintiffs might ask the district court to apply the more forgiving Cook IX standard in further proceedings on remand and quite rightly sought to discourage that possibility, avoid fifteen more years of squabbling, and inch the case toward resolution by clarifying the guiding principles for remand on a range of issues the parties briefed to the court.
So we find ourselves looping back to where we began. This court never “expressly disapproved” the district court’s Colorado law nuisance instructions. To the contrary, in the first appeal the .defendants never meaningfully challenged the jury’s instructions on state nuisance law and, in any event, this court expressly approved them. The only thing disapproved in the first appeal was an inoperative pretrial order. Neither did Dow and Rockwell dispute that the evidence presented at trial sufficed to support a common law nuisance verdict under the legally correct instructions the district court issued to the jury. So when it came to a state law nuisance claim, by the end of the first appeal there existed a properly instructed jury, legally sufficient evidence, and a favorable jury verdict — elements that usually guarantee a judgment for the plaintiffs, not preclude one. See Lloyd v. Grynberg, 464 F.2d 622, 625 (10th Cir.1972) (“The law is clear that an appellate court will not upset or disturb a jury verdict if the case has been submitted upon proper and adequate instructions and the verdict is supported by the evidence.”).
B
At this point the defendants hold but one more card to play. They suggest that, whatever the first panel’s intentions may have been, the language it used in its mandate barred the district court as a formal matter from entering a state law nuisance judgment on the basis of the existing jury verdict. Of course, the defendants can’t offer any reason why the first panel’s mandate should have operated this way. For example, they (again) don’t contend that the prior panel considered the preemption question we’ve addressed and rejected the analysis we’ve given. And by this point they can’t contend that the first panel found error in the district court’s nuisance instructions or any other aspect of the nuisance verdict. So it is that the defendants ask us to suppose that the mandate precluded entry of a nuisance judgment on remand for no good reason, accidentally even.
We decline this invitation. “The scope of the mandate ... is carved out by exclusion: unless the district court’s discretion is specifically cabined, it may exercise discretion on what may be heard.” Dish Network Corp. v. Arrowood Indem. Co., 772 F.3d 856, 864 (10th Cir.2014) (brackets omitted). And examining the mandate from the first appeal, we see nothing that “specifically” precluded the district court from entering a new judgment on a common law nuisance claim on the basis of the existing jury verdict. Rather, after rejecting the district court’s understanding of what qualifies as a nuclear incident for purposes of the Price-Anderson Act, the panel’s opinion ended this way: “[TJhis court REVERSES and REMANDS the *1102case to the district, court. We DIRECT the district court to vacate the judgment and conduct further proceedings not inconsistent with this opinion.” Cook, 618 F.3d at 1153. It opened with identical language. See id. at 1133. By its plain terms, this language prevented the district court from entering a new judgment predicated on any erroneous understanding of the Price-Anderson Act. But beyond that the mandate didn’t specifically prevent .anything. Indeed, if anything its use of the double negative suggested that the district court was left with maximum room to maneuver on remand — expressly allowing the district court to undertake any proceedings “not inconsistent” with this court’s opinion.
Nothing about this should come as a surprise. After a court of appeals vacates a judgment a district court on remand often will enter a new judgment in the same party’s favor on the existing record if one can be had unaffected by the error found in the appeal. That’s what happened in Dish Network when the district court entered a new order granting summary judgment after this court considered and rejected a previous order granting summary judgment on different grounds. See 772 F.3d at 865-66. That’s what happens all the time after an appellate court reverses a judgment following a jury trial because of an error affecting one claim but not others and the district court proceeds to fashion a new judgment using only the portions of the existing verdict unaffected by that error.5 We discern no reason why a different result should obtain here, where a judgment implicating the Act’s protections is no longer possible but all the elements necessary for a judgment under state law are present and unaffected by any error identified in the first appeal, or on remand for that matter.
One wrinkle remains — though it’s a wrinkle the defendants don’t pursue and thus have waived any complaint about. See United States v. Gama-Bastidas, 222 F.3d 779, 784 (10th Cir.2000) (discussing the non-jurisdictional nature of the mandate rule). In one spot in the middle of its opinion the first panel explained its intentions a little differently, writing that “[b]e-cause the jury was not properly instructed on an essential element of Plaintiffs’ PAA claims [the definition of ‘nuclear incident’], the verdict must be set aside and the case remanded for further proceedings not inconsistent with this opinion.” Cook, 618 F.3d at 1142. Here alone the panel spoke of setting aside a verdict rather than just the judgment. But even if the defendants had pursued an argument based on this language, it’s plain to us that the panel didn’t mean to preclude the entry of a new judgment based on the existing state law nuisance verdict. The panel referred here only to the plaintiffs’ “PAA claims” and identified only an error of federal, not state, law (the lack of a “proper[] in*1103struct[ion]” about the term “nuclear incident”). Id. There is no suggestion in this discussion or anywhere of an error in the state law aspects of 'the jury’s verdict. Indeed, our understanding on this score is reinforced by the context in which this sentence appears: at the end and as a summary of the panel’s discussion of the Price-Anderson Act and its nuclear incident requirement. This discussion is separate from the specifically captioned discussion of the plaintiffs’ state law nuisance allegations that endorsed the district court’s jury instructions. Finally, our conclusion is cemented by the language at either end of the opinion that makes no mention of any error in the nuisance verdict and discusses only a reversal of the judgment. In discerning the terms of an earlier court’s mandate, some courts examine the whole of the opinion and some place particular stress on any language at the end or beginning summarizing the court’s instructions.6 Either way, all roads in this case lead to the same place: a conclusion that the first panel did not specifically preclude the district court from entering a new judgment predicated on an error-free state law nuisance verdict.
C
Our concurring colleague sees a few things differently but, happily, we agree on the important things. We agree that the Price-Anders.on Act does not preempt and preclude a freestanding state law nuisance claim when a nuclear incident is alleged but unproven. See supra Part II.B. We agree that the defendants’ primary mandate argument — suggesting that the first panel expressly disapproved the plaintiffs’ state law nuisance theory — is mistaken. See supra Part III.A. We even agree that the first panel’s mandate, as a formal matter, can’t be read to preclude the entry of a freestanding state law nuisance judgment. See supra Part III.B. To be sure, our colleague would remand the case for another trial on the plaintiffs’ state law nuisance claim rather than permit the entry of a judgment now based on the existing nuisance verdict. But that’s a pretty small distance between us. It’s a distance, too, that stems only from slight differences in how we read the first panel’s mandate: as we’ve explained we see nothing in the mandate specifically precluding entry of a judgment based on the existing nuisance verdict; the concurrence does, though it reads the mandate as allowing a second trial on the same claim.
Unfortunately, we cannot close that small gap remaining between us. For one thing, the concurrence doesn’t explain how we might vacate the district court’s judgment and send the case back for a new trial when the plaintiffs didn’t ask for a new trial on remand and don’t now. Neither do we see how we might remand for a new trial if, as the concurrence suggests, the plaintiffs’ briefing in this appeal waived any complaint about the district court’s mandate ruling. After all, that ruling held the plaintiffs’ state law claims not “pursua-ble” and, if unchallenged, it would seem to supply a basis for affirmance, not reversal. For our part, of course, we do not believe that the plaintiffs waived any complaint about the district court’s mandate ruling— indeed, the defendants have never suggested as much, and we are not inclined to pursue an argument for the defendants that they have not pursued for themselves. *1104See Cook, 618 F.3d at 1138-39. Finally, we just don’t see how the first panel’s mandate “specifically” prohibited the entry of a judgment based on the existing nuisance verdict. Of course, the concurrence points to the wrinkle we’ve just discussed-language in' the middle of the opinion suggesting the possibility of setting aside the jury’s verdict, not just the judgment. But here again the concurrence would have us rest on a forfeitable argument that Dow and Rockwell haven’t presented. And it would have us focus on language from the middle of an opinion— without reference to the language at the opinion’s beginning or end, and without reference even to language in the surrounding paragraph that discusses errors only in the application of federal, not state, law. While we are aware of different views on the question whether an appellate mandate should be read to include only the formal instructional language at the beginning and end of an opinion or should be discerned in light of the whole of an opinion, see supra at 1103 n. 6, we are aware of no authority suggesting we should focus only on isolated language in the middle of an opinion.7
IV
In the end, Dow and Rockwell appear to have persuaded even the plaintiffs that this case does not involve a nuclear incident within the meaning of the Price-Anderson Act — at least in light of the statutory construction the defendants urged and this court adopted in the first appeal. But that does not mean the defendants are insulated from any liability — or that the jury’s verdict is a pointless piece of paper. In two separate appeals spanning many years the defendants have identified no lawful impediment to the entry of a state law nuisance judgment on the existing verdict. They have shown no preemption by federal law, no error in the state law nuisance instructions, no mandate language specifically precluding this course. No other error of any kind is even now alleged.
*1105In this light, we can well understand why the plaintiffs on remand renounced a new trial and sought entry of a judgment based on the existing nuisance verdict. Indeed, without some specific mandate or identified error requiring so much we can imagine only injustice flowing from any effort to gin up the machinery of trial for a second pass over terrain it took fifteen years for the first trial to mow through. Injustice not only in the needless financial expense and the waste of judicial resources, but injustice in the human costs associated with trying to piece together faded memories and long since filed away evidence, the emotional ordeal parties and witnesses must endure in any retrial, the waste of the work already performed by a diligent and properly instructed jury, and the waiting — the waiting everyone would have to endure for a final result in a case where everyone’s already waited too long, longer even than the lives of today’s college graduates. When the district court receives this case it should proceed to judgment on the existing nuisance verdict promptly, consistent with resolving the outstanding class action question, wary of arguments that have already been rejected or forfeited. This long lingering litigation deserves to find resolution soon. The judgment of the district court is vacated and the case is remanded for proceedings consistent with this opinion.
. The defendants didn't (and don’t) challenge this move, and surely would be judicially es-topped from doing so in any event. In the last appeal, they claimed that the plaintiffs couldn't prove a nuclear incident given the governing law and the facts of this case. Appellants' Supplemental Br. at 13, Cook, 618 F.3d 1127 (Nos.08-1224, 08-1226, 08-1239) (''[P]laintiffs did not and could not prove their claims of a classwide ‘nuclear incident.' ”). So it would be pretty difficult for them to argue otherwise now: "[Ajbsent any good explanation, a party should not be allowed to gain an advantage by litigation on one theory, and then seek an inconsistent advantage by pursuing an incompatible theory.” New Hampshire v. Maine, 532 U.S. 742, 749, 121 S.Ct. 1808, 149 L.Ed.2d 968 (2001) (quoting 18 Charles Alan Wright et al., Federal Practice and Procedure § 4477, at 782 (1981)).
. Neither do we doubt that subject matter jurisdiction would exist to support a state law nuisance judgment, even supposing the Price-Anderson Act’s grant of federal question jurisdiction dissipated when the plaintiffs gave up their effort to assert or prove a nuclear incident. After all, the complaint alleged that the parties were completely diverse and asked for damages in excess of the statutory minimum. See 28 U.S.C. § 1332 (1988). Likewise, the district court retained pendent jurisdiction to consider the state law nuisance claim given the advanced stage of this litigation and the resources it had already consumed. See United Mine Workers of Am. v. Gibbs, 383 U.S. 715, 725, 86 S.Ct. 1130, 16 L.Ed.2d 218 (1966); Russillo v. Scarborough, 935 F.2d 1167, 1172 n. 5 (10th Cir.1991).
. If we were to adopt the defendants' interpretation of the Act as barring any recovery in cases involving lesser nuclear occurrences, the plaintiffs contend we would encounter a serious constitutional problem. As a matter of due process, they argue, Congress cannot eliminate longstanding common law rights without providing any "reasonable alternative remedy” unless there is a "compelling” reason to do so. PruneYard Shopping Ctr. v. Robins, 447 U.S. 74, 93-94, 100 S.Ct. 2035, 64 L.Ed.2d 741 (1980) (Marshall, J., concurring). This is of course no trivial argument, but because we find the defendants’ reading of the statute unconvincing on its own terms it is also one we happily need not reach or resolve.
. The only instruction that asked the jury to assess individual plaintiffs’ subjective reactions was Instruction 3.28, which had nothing to do with the nuisance instructions that appeared in Instructions 3.6 through 3.17. If anything, its appearance outside the context of the court’s nuisance instructions serves to underscore its absence inside that context.
. See, e.g., United States v. Hicks, 146 F.3d 1198, 1200-01 (10th Cir.1998) (affirming district court’s entry of judgment on unaffected counts after remand without necessity of new trial); Reeder-Simco GMC, Inc. v. Volvo GM Heavy Truck Corp., 497 F.3d 805, 807 (8th Cir.2007) (recounting district court's entry of modified judgment on remand on unaffected claims following reversal by Supreme Court); O'Hagan v. Soto, 565 F.Supp. 422, 425-26, 429 (S.D.N.Y.1983) (noting that initial judgment as to question of liability survived subsequent vacaturs on question of damages); Ponte v. Real, 471 U.S. 491, 503 n. 4, 105 S.Ct. 2192, 85 L.Ed.2d 553 (1985) (Stevens, J., concurring in part) (collecting examples where judgment on federal claim is reversed but state law judgments are permitted to stand); see also Harvey v. Richards, 11 F. Cas. 740, 745 (C.C.D.Mass.1814) (Story, J.) ("At common law, if a plaintiff obtain a judgment in an inferior tribunal, which is reversed in the appellate court, it is very clear, that the reversal operates no further, than to nullify the original judgment.”)
. Compare, e.g., Apponi v. Sunshine Biscuits, Inc., 809 F.2d 1210, 1217 (6th Cir.1987) (looking to "the entire opinion of the court”), and Procter & Gamble Co. v. Haugen, 317 F.3d 1121, 1126 (10th Cir.2003) (same), with, e.g., Hicks v. Gates Rubber Co., 928 F.2d 966, 968-69 (10th Cir.1991) (focusing on the final paragraph), and Gibson v. Worley Mills, Inc., 620 F.2d 567, 567 (5th Cir.1980) (per curiam) (referring to an opinion's "last paragraph” as "the mandate”).
. Even if the first panel opinion contains no express mandate requiring a new trial, the concurrence suggests we should draw the same message by inference from indirect evidence. But as we’ve seen, the mandate rule operates by exclusion and any avenue a court of appeals does not specifically foreclose remains available on remand. See supra at 1101-02. Beyond that, we are unable to agree with the inferences the concurrence draws from the indirect evidence it cites. The concurrence emphasizes the plaintiffs’ repeated request that we 'reinstate the verdict,” arguing that it shows even the plaintiffs recognize our prior decision as having overturned the verdict on their state law nuisance claim. But the plaintiffs have consistently claimed that no error identified in the first appeal touched their state law claim, that nothing in the Act preempts that claim, and that there’s nothing stopping a court from entering judgment on it now. We agree that it would have been more accurate to call this an argument for “reinstating the judgment based on an error-free verdict” rather than "reinstating the verdict.” But the plaintiffs' imprecise wording surely cannot alter the scope of this court’s mandate (or even obscure the nature of their remedial request). Neither are we able to agree with the significance the concurrence attaches to the first panel’s decision not to address certain claims of error because they might not arise in the event of a new trial. We don't doubt it’s possible the prior panel thought it likely the plaintiffs would want to seek the Act’s benefits on remand and would pursue a new trial under the Act. But a mandate isn't governed by what's in someone’s head; it’s governed by what’s on the page. So an appellate court’s inability to imagine every tactical decision a party may make months or years later on remand does not dictate the scope of the mandate written in its opinion. Neither, of course, did the first panel’s decision to defer certain issues preclude the defendants from raising them on remand if they really thought them sufficient to bar entry of judgment on the existing nuisance verdict.