BIBAS, Circuit Judge. Cleaning up pollution is both essential and expensive. But polluters often frustrate cleanups, blaming one another to shift their fair share of the costs. The Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) tries to fix this problem by forcing them all to pay: A government or person who incurs cleanup costs can sue a site’s polluters to recover those costs in a “cost recovery” action. If the plaintiff prevails, the polluters are held strictly as well as jointly and severally liable. Then, the polluters can ask a court to split the bill among them equitably in a “contribution” action. If a polluter first settles its liability with the government, CERCLA immunizes that polluter from contribution liability to other polluters. But this immunity comes with a cost: a pol- luter who has settled with the government may bring a contri- bution claim against other polluters, but not a cost-recovery claim. Appellant Cranbury Brick Yard, LLC, brings both cost- recovery and contribution claims against the federal govern- ment. It seeks to recover money that it spent cleaning up a long- abandoned weapons-manufacturing facility that the U.S. mili- tary and others contaminated. But Cranbury Brick Yard settled its potential CERCLA liability with the State of New Jersey before the cleanup. That gave it immunity from contribution claims, which extinguishes its cost-recovery claim. This left only its contribution claim against the federal government. But that claim is untimely because Cranbury Brick Yard sued nine years after joining the settlement. So we will affirm. I. CERCLA, BRIEFLY EXPLAINED CERCLA is “notorious for its lack of clarity and poor draftsmanship.” Lansford-Coaldale Joint Water Auth. v.

Tonolli Corp., 4 F.3d 1209, 1221 (3d Cir. 1993); accord Gio- vanni v. U.S. Dep’t of the Navy, 906 F.3d 94, 117 (3d Cir. 2018) (“CERCLA is not the Mona Lisa of statutes.”). Still, its broad contours are easy enough to grasp. CERCLA gives private litigants two causes of action: cost recovery under section 107(a) and contribution under sec- tion 113(f)(1). 42 U.S.C. §§ 9607(a), 9613(f)(1). The two are distinct, but they “complement each other” by applying “ ‘to persons in different procedural circumstances.’ ” United States v. Atl. Research Corp., 551 U.S. 128, 139 (2007) (quot- ing Consol. Edison Co. of N.Y. v. UGI Utils., Inc., 423 F.3d 90, 99 (2d Cir. 2005)); see Cooper Indus., Inc. v. Aviall Servs., Inc., 543 U.S. 157, 163 n.3 (2004). We discuss each in turn. A. Cost recovery Anyone who cleans up a brownfield may sue the site’s pol- luters to recover “any . . . necessary costs of response.” 42 U.S.C. § 9607(a)(4)(B). If a cost-recovery suit succeeds, the defendants are strictly as well as jointly and severally liable. Pa. Dep’t of Envtl. Prot. v. Trainer Custom Chem., LLC, 906 F.3d 85, 89–90 (3d Cir. 2018). For “remedial action[s]” (that is, “those actions consistent with permanent remedy”), the stat- ute of limitations for a cost-recovery action is six years from when the cleanup begins. 42 U.S.C. §§ 9601(24), 9613(g)(2)(B). Four kinds of “potentially responsible part[ies]” may be li- able for cost recovery, including the site’s current owner and anyone who owned the site “at the time of disposal of any haz- ardous substance.” Id. § 9607(a)(1)–(2); 40 C.F.R. § 304.12(m). But an owner is immune from liability if it bought the site as a “bona fide prospective purchaser.” 42 U.S.C. §§ 9601(40), 9607(r)(1). To qualify, the owner must show (among other things) that “[a]ll disposal of hazardous substances . . . occurred before [it] acquired the facility.” Id. § 9601(40)(B)(i). B. Contribution If a polluter is or may be liable under CERCLA or has set- tled its liability with a state or the federal government, it may sue other polluters for “contribution.” Id. § 9613(f)(1), (3)(B). Contribution is a “tortfeasor’s right to collect from others re- sponsible for the same tort after the tortfeasor has paid more than his or her proportionate share, [with] the shares being de- termined as a percentage of fault.” Atl. Research, 551 U.S. at 138 (quoting Contribution, Black’s Law Dictionary 353 (8th ed. 2004)). So a contribution action lets a court “allocate re- sponse costs among liable parties using . . . equitable factors.” 42 U.S.C. § 9613(f)(1). C. Resolving the overlap Cost recovery and contribution are “similar and somewhat overlapping remed[ies].” Key Tronic Corp. v. United States, 511 U.S. 809, 816 (1994). That overlap became an issue after the Supreme Court’s decision in Atlantic Research, which held that polluters could bring both kinds of claims against one an- other. 551 U.S. at 141. Since then, we and our sister circuits have clarified the relationship between the two kinds of claims. 1. The price of contribution-claim immunity. A polluter who settles its CERCLA liability with the federal government or a state government enjoys immunity under § 9613(f)(2) from contribution claims. In Agere Systems, we held that if a polluter is immune from contribution claims, it cannot bring cost- recovery claims. Agere Sys., Inc. v. Advanced Envtl. Tech. Corp., 602 F.3d 204, 229 (3d Cir. 2010). Instead, it can bring only contribution claims. Id. Six other circuits have reached this issue. All agree. See Whittaker Corp. v. United States, 825 F.3d 1002, 1007 & n.4 (9th Cir. 2016); Hobart Corp. v. Waste Mgmt. of Ohio, Inc., 758 F.3d 757, 768 (6th Cir. 2014); Bernstein v. Bankert, 733 F.3d 190, 202 (7th Cir. 2013); Solutia, Inc. v. McWane, Inc., 672 F.3d 1230, 1236–37 (11th Cir. 2012) (per curiam); Morri- son Enters. v. Dravo Corp., 638 F.3d 594, 603 (8th Cir. 2011); Niagara Mohawk Power Corp. v. Chevron U.S.A., Inc., 596 F.3d 112, 127–28 (2d Cir. 2010). 2. Why the Agere Systems rule is necessary. An example may help show (a) the normal interplay between cost-recovery and contribution claims, (b) why contribution-claim immunity under § 9613(f)(2) encourages CERCLA settlements, and (c) how the Agere Systems rule helps apportion cleanup costs equitably by keeping polluters from using their immunity as both sword and shield. Imagine a cleanup of comic proportions: Years ago, Wayne Enterprises and LexCorp ran a chemical factory together. After it shut its doors, the site remained contaminated with hazardous substances. Now, Gotham City owns the vacant, polluted brownfield. Responding to mounting public pressure, the City spends $1 billion cleaning up the site. Then it sues both com- panies for cost recovery.

In a typical CERCLA lawsuit, this is how cost-recovery and contribution claims would interact: because cost-recovery ac- tions impose joint-and-several liability, both companies would theoretically be on the hook for 100% of the billion-dollar lia- bility. Trainer Custom, 906 F.3d at 90. While joint-and-several liability helps compensate fully those who clean up pollution, it sometimes leads to inequitable results. Fortunately, CERCLA has an answer: the two companies could cross-claim for contribution and ask the court to split the bill between them in proportion to their percentages of fault. See Atl. Research, 551 U.S. at 139–40 (citing 42 U.S.C. § 9613(f)(1)). But litigation is costly and time consuming. So CERCLA encourages polluters to settle with the government to get clean- ups started sooner. Contribution-claim immunity under § 9613(f)(2) promotes those efficient settlements. Imagine in- stead that the Environmental Protection Agency starts an en- forcement action, claiming that LexCorp is responsible for $400 million of the cleanup costs at the Gotham chemical fac- tory. LexCorp’s lawyers negotiate down and settle for just $100 million. Afterwards, Wayne Enterprises cannot bring a contribution action and ask the court to allot more than $100 million in cleanup costs to LexCorp. If it could, it would undermine the finality of CERCLA settlements and thus dis- courage them. Cf. Atl. Research, 551 U.S. at 141 (noting that “settlement carries the inherent benefit of finally resolving lia- bility”). Instead, contribution-claim immunity under § 9613(f)(2) protects settlements like LexCorp’s. Without the Agere Systems rule, however, settling polluters could wield this immunity offensively to escape liability. If LexCorp settles with the government for $100 million and then sues Wayne Enterprises for cost recovery, LexCorp could im- pose 100% of its liability on Wayne Enterprises through joint- and-several liability. Agere Sys., 602 F.3d at 228. Worse still, because LexCorp has § 9613(f)(2) immunity, Wayne Enter- prises could not counterclaim for contribution to apportion costs equitably. So LexCorp could recoup its $100 million in settlement costs even though it is “actually responsible for, and ha[s] stipulated that [it is] responsible for, a significant portion of the contamination.” Id. at 228. So if a polluter like LexCorp who is immune from contri- bution claims could bring a cost-recovery action, it could use that immunity to “ ‘eschew equitable apportionment,’ ” impose “the harshness of joint and several liability” against other pol- luters, and shift its own liability in the process. Id. at 228–29 (quoting Atl. Research, 551 U.S. at 138). To avoid this “per- verse result,” we held in Agere Systems that a polluter who has contribution-claim immunity under § 9613(f)(2) cannot bring a cost-recovery claim. Id. Instead, it can bring only a contribu- tion claim. Id. at 229. II. FACTUAL BACKGROUND A. From bombs to brownfield 1. The site’s troubled history. During World War II and the Korean War, Unexcelled Manufacturing Co. operated a weap- ons factory in Cranbury, New Jersey. It made bombs, anti- aircraft ammunition, grenade fuses, and other high-powered weapons for the U.S. military.

In 1954, a warehouse at the site containing hundreds of thousands of grenade fuses exploded. Two people died and several were injured. Though the military sent soldiers from nearby Fort Dix to help with the cleanup, lots of hazardous ma- terials were left, including unexploded weapons and dangerous chemicals. Soon after, the factory shut down. But the pollution remained. 2. New Jersey’s enforcement action. Twenty years later, Cranbury Development Corp. bought the site. Thirty years af- ter that, the New Jersey Department of Environmental Protec- tion (NJDEP) finished an investigation of the site. It issued a directive identifying the parties responsible for the site’s con- tamination: Cranbury Development (the site’s owner), Maxxam Group, Inc. (the successor to Unexcelled Manufac- turing), and the U.S. Navy (which had funded and effectively controlled the site). NJDEP directed them to “memorialize their commitment to perform the remediation in an Adminis- trative Consent Order.” App. 139. But the Navy rebuffed NJDEP and refused to take part. In 2005, Cranbury Development and Maxxam entered into a Consent Order with NJDEP. The Consent Order did three things: First, Cranbury Development and Maxxam agreed to clean up the site. Second, NJDEP agreed not to sue them if they complied. Third, all the parties agreed that the “Consent Order constitutes an administrative settlement within the meaning of CERCLA” and “resolve[s] the liability of [Cranbury Develop- ment] and Maxxam to the State of New Jersey for some or all of” the cleanup costs. App. 147. That settlement with NJDEP let Cranbury Development and Maxxam seek contribution from other polluters (like the Navy), while immunizing them from such claims. 42 U.S.C. § 9613(f)(2). 3. Cranbury Brick Yard buys the site and joins the Consent Order. In early 2006, Cranbury Brick Yard bought the site from Cranbury Development with plans to redevelop it into commercial warehouses. But first, it sought an agreement with NJDEP to assume Cranbury Development’s cleanup obliga- tions. Rather than strike its own deal with NJDEP, Cranbury Brick Yard agreed to join the existing agreement. So the Con- sent Order was amended to “remove Cranbury Development” and “replace” it with Cranbury Brick Yard “as a Respondent.” App. 159. That substitution applied throughout the whole orig- inal agreement, which “remain[ed] in full force and effect.” App. 160. NJDEP approved the amendment, to which the par- ties attached the original agreement. By joining the Consent Order, Cranbury Brick Yard, like Cranbury Development, agreed to clean up the site and settle its potential CERCLA liability to NJDEP. In return, it too gained immunity from contribution claims. 42 U.S.C. § 9613(f)(2). 4. The cleanup runs into problems. After joining the Con- sent Order, Cranbury Brick Yard inspected the site thoroughly and developed a cleanup plan. In 2013, after Cranbury Brick Yard got the permits it needed, its contractors began cleaning up the site.

But during the cleanup, problems arose. Among them, the contractors unexpectedly punctured and dug up an under- ground tank holding water and roughly twenty gallons of pe- troleum. Some of the liquid spilled. Cranbury Brick Yard promptly notified NJDEP. Then it mixed the contaminated dirt with clean dirt and reburied it on-site. Despite this and other bumps in the road, Cranbury Brick Yard kept cleaning up the site. It claims to have spent well over $50 million in the process. B. Cranbury Brick Yard sues the federal government In April 2015, nine years after joining the Consent Order but just over two years after it began the cleanup, Cranbury Brick Yard sued the federal government. It claimed that the government was responsible for some of the cleanup costs be- cause the military was partly responsible for the site’s contam- ination. It sought both cost recovery and contribution under CERCLA. 42 U.S.C. §§ 9607(a), 9613(f)(1). The government unsuccessfully moved to dismiss. It then counterclaimed for contribution, alleging that Cranbury Brick Yard was also responsible for some of the site’s pollution. In response, Cranbury Brick Yard raised an affirmative defense: even if it were liable, the amended Consent Order would im- munize it from contribution liability. After discovery, both sides moved for summary judgment. The District Court granted the government’s motion and de- nied Cranbury Brick Yard’s. Cranbury Brick Yard, LLC v. United States, No. 3-15-cv-02789, 2018 WL 4828410, at *1 (D.N.J. Oct. 3, 2018). It reached three holdings.

First, the District Court held that the amended Consent Or- der was a settlement under § 9613(f)(2), so it immunized Cran- bury Brick Yard from contribution liability. 2018 WL 4828410, at *4. Under our holding in Agere Systems, this meant that Cranbury Brick Yard could not bring a cost- recovery claim. Id. at *4–5 (citing Agere Sys., 602 F.3d at 229). Second, it held that Cranbury Brick Yard brought its re- maining contribution claim too late. Id. at *6. It reasoned that the “applicable statute of limitations is three years” from “when all parties to the [Consent Order] Amendment . . . signed it.” Id. Because Cranbury Brick Yard sued nine years later, its contribution claim was untimely. Id. at *5–6. Third, it rejected Cranbury Brick Yard’s argument that it had immunity as a “bona fide prospective purchaser.” Id. at *7; see 42 U.S.C. §§ 9601(40), 9607(r)(1). The District Court found that reburying the contaminated dirt on-site amounted to a “disposal.” 2018 WL 4828410, at *6–7 (citing 42 U.S.C. § 9601(40)(B)(i)); see 42 U.S.C. §§ 6903(3), 9601(29) (defin- ing “disposal”). And because this alleged disposal happened after Cranbury Brick Yard acquired the site, the District Court determined that it had lost its status as a bona fide prospective purchaser. 2018 WL 4828410, at *7. Cranbury Brick Yard appeals all three holdings. The Dis- trict Court had jurisdiction under 28 U.S.C. § 1331, and we have jurisdiction under § 1291. We review the grant or denial of summary judgment de novo. Tundo v. County of Passaic, 923 F.3d 283, 286 (3d Cir. 2019).

III. THE AMENDED CONSENT ORDER EXTINGUISHED CRANBURY BRICK YARD’S COST-RECOVERY CLAIM

A. Cranbury Brick Yard has contribution-claim im- munity, so it cannot bring a cost-recovery claim

The text of the amended Consent Order is clear. It did four things:

• Substitution: The amendment “remove[d]” Cranbury Development and “replace[d]” it with Cranbury Brick Yard as a party to the whole Consent Order. App. 159.

• Incorporation: The amendment “bec[a]me part of” the original Consent Order, which “remain[ed] in full force and effect” and was attached to the amendment. App. 160.

• Settlement: The amendment recognized Cranbury Brick Yard’s potential “liab[ility] to the State of New Jersey under CERCLA” and “resolve[d]” it in exchange for Cranbury Brick Yard assuming “responsib[ility] for re- mediation of the Site.” App. 161, 170.

• Immunity: The amendment “constitute[d] an adminis- trative settlement within the meaning of CERCLA.” App. 170. It explicitly recognized Cranbury Brick Yard’s contribution-claim immunity under § 9613(f)(2). Id. And it gave Cranbury Brick Yard “the right to seek contribution” from other polluters under § 9613(f)(3)(B). Id.

The effect of its language is also clear. Because the amended Consent Order gave Cranbury Brick Yard immunity from contribution claims, it barred Cranbury Brick Yard from seeking cost recovery from other potentially responsible par- ties. Agere Sys., 602 F.3d at 229. So the District Court rightly granted summary judgment for the government, cutting off Cranbury Brick Yard’s cost-recovery claim. B. Cranbury Brick Yard’s counterarguments lack merit

Despite this language, Cranbury Brick Yard argues that it still has a viable cost-recovery claim. It does not. 1. The voluntariness of the cleanup is no defense. Cranbury Brick Yard first contends that a party who “voluntarily cleans up a contaminated site and incurs its own response costs” must be able to seek cost recovery. Appellant’s Br. 29. Not so. Agere Systems’s bar on cost-recovery claims applies as long as a set- tlement makes that party immune from contribution claims. 602 F.3d at 229. Voluntariness is irrelevant. In any event, Cranbury Brick Yard’s cleanup costs were not fully voluntary. True, it joined the amended Consent Order freely. But after that, the agreement required it to clean up the site. 2. Cranbury Brick Yard benefited from settling its poten- tial liability. Next, Cranbury Brick Yard claims that the amended Consent Order is not a settlement under § 9613(f)(2). It argues that it had no liability to settle and thus did not have to negotiate a settlement. Appellant’s Br. 31, 33–35.

But Cranbury Brick Yard had at least two reasons to settle. First, it did have something to settle: its potential liability as the site’s current owner. See 42 U.S.C. § 9607(a)(1). Second, the settlement gave Cranbury Brick Yard valuable immunity from contribution claims. Id. § 9613(f)(2). Indeed, that immun- ity proved useful when Cranbury Brick Yard pleaded it as an affirmative defense to the government’s contribution counter- claim. Cranbury Brick Yard was correct then and is mistaken now. Cranbury Brick Yard also argues that the District Court “in- correctly assumed” that the parties intended to extend Cran- bury Development’s and Maxxam’s settlement to Cranbury Brick Yard. Appellant’s Br. 33. But the Court never “assumed” the parties’ intent—it knew their intent from the unambiguous text of the amended agreement. 3. Cranbury Brick Yard’s other agreements are irrelevant. The text of the amended Consent Order is clear. Thus, we reject Cranbury Brick Yard’s resort to parol evidence. In any event, this evidence is unconvincing. In a separate Purchase and Sale Agreement for the site, Cranbury Brick Yard agreed to assume only Cranbury Development’s “obliga- tions” under the Consent Order. App. 324. But “obligation” in- cludes “anything that a person is bound to do or forbear from doing.” Obligation, Black’s Law Dictionary 1292 (11th ed. 2019). So that agreement tracks the amended Consent Order: both require that Cranbury Brick Yard assume the benefits and burdens of the original Consent Order.

4. A different oversight document would have sufficed. Cranbury Brick Yard also claims that it had to join the Consent Order because New Jersey regulations then in effect required an “applicable oversight document” before it could clean up the site. Appellant’s Br. 32 (quoting N.J. Admin. Code § 7:26C-1.1 (2006)). But it concedes that the amended Consent Order was just “[o]ne such oversight document.” Id. It could have signed any other agreement that “define[d] the role of [the company] participating in the remediation of a contaminated site.” See N.J. Admin. Code § 7:26C-1.3 (2006) (definition of “[o]versight document”). It did not. Instead, it joined the amended Consent Order, which gave it contribution-claim im- munity. And doing so extinguished its cost-recovery claim. 5. Cranbury Brick Yard forfeited its argument about the federal government’s model order. Finally, Cranbury Brick Yard claims that a “model” consent order written by two fed- eral agencies supports its cost-recovery claim. But Cranbury Brick Yard forfeited this argument by failing to make it in the District Court. We see no exceptional circumstances that might cause us to overlook this forfeiture. In any event, this “model” was issued nearly a year after Cranbury Brick Yard joined the amended Consent Order. So while it might provide guidance to the next company in Cranbury Brick Yard’s situation, it casts no light on how we should read this Consent Order. IV. CRANBURY BRICK YARD’S CONTRIBUTION CLAIM IS UNTIMELY

The District Court properly granted summary judgment for the government on Cranbury Brick Yard’s cost-recovery claim. Now we must decide whether its remaining claim for contribution is timely. It is not. So we will affirm the District Court on this issue too. A. The clock starts ticking when a party administra- tively settles its liability

1. CERCLA’s statutes of limitations. CERCLA has two different statutes of limitations: one for cost-recovery claims and another for contribution claims. 42 U.S.C. § 9613(g)(2)– (3). Section 9613(g)(2)(B) provides that a “remedial” cost- recovery action like the one that Cranbury Brick Yard brought here must be filed “within 6 years after initiation of” the cleanup. Section 9613(g)(3) provides that “[n]o action for contribu- tion . . . may be commenced more than [three] years after” one of four triggering events: (1) “the date of judgment in any ac- tion” for “recovery of . . . costs or damages”; (2) the date of a “de minimis” settlement under § 9622(g); (3) the date of a cost- recovery settlement with the federal government under § 9622(h); or (4) the date of “entry of a judicially approved set- tlement with respect to . . . costs or damages.” Id. § 9613(g)(3) (emphasis added). 2. We must fill a gap in the statute of limitations. The par- ties agree that none of these contribution-claim triggers oc- curred because the Consent Order was not “judicially ap- proved.” Looking only at these provisions, one might think that Cranbury Brick Yard’s contribution claim has no time limit. So the statute of limitations has a gap. But it is not unusual for federal statutes to have such gaps. See N. Star Steel Co. v. Thomas, 515 U.S. 29, 33 (1995). We usually fill these gaps by borrowing state statutes of limita- tions. See id. at 33–34. But doing so with CERCLA would sub- ject the federal government to varying liability state by state. See 42 U.S.C. § 9620(a)(1) (waiving sovereign immunity); see also FMC Corp. v. U.S. Dep’t of Commerce, 29 F.3d 833, 840 (3d Cir. 1994) (en banc) (under CERCLA, “when the govern- ment engages in activities that would make a private party lia- ble if the private party engaged in those types of activities, then the government is also liable” (emphasis omitted)). Plus, CERCLA provides ready alternatives. We may bor- row a federal limitations period where “federal law clearly pro- vides a closer analogy than available state statutes, and when the federal policies at stake . . . make that rule . . . significantly more appropriate.” DelCostello v. Int’l Bhd. of Teamsters, 462 U.S. 151, 172 (1983). That is true here. To fill the gap, we could hold, as the Sixth Circuit has, that § 9613(g)(3)’s list is illustrative, not exhaustive. See Hobart Corp., 758 F.3d at 774–75 (declining to read the “enumeration of a few triggering events to preclude finding others”). If we did that, then we could expand the list to encompass settle- ments that have not been judicially approved. See id. But this approach is hard to square with § 9613(g)(3)’s text and structure. That paragraph lists just four events that trigger finality. It separates each with the word “or.” And it contains no broad residual language, like “or other similar settlements.” See Circuit City Stores, Inc. v. Adams, 532 U.S. 105, 114–15 (2001); see also Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 202 (2012) (noting that ejusdem generis applies only where there is “broad catchall language”). So we decline to treat the statutory list as illustra- tive. Instead, we fill the gap by “ ‘borrow[ing]’ the most suitable statute or other rule of timeliness.” DelCostello, 462 U.S. at 158. Here, the “most suitable” rule comes from § 9613. 3. The clock starts ticking when a party administratively settles its liability. The thrust of § 9613(g)(3) is that the limita- tions period starts to run when a litigant’s contribution claim accrues—when it “come[s] into existence as an enforceable claim or right.” Accrue, Black’s Law Dictionary 26 (11th ed. 2019). Three textual clues show that a contribution claim ac- crues when a litigant’s potential CERCLA liability is formally recognized. First, § 9613(f)(1) lets a litigant sue for contribution “during or following any [cost-recovery] action.” So a litigant can bring a contribution claim after its possible liability is officially recognized in a complaint or judgment. Second, § 9613(f)(3)(B) states that “[a] person who has resolved its lia- bility to the United States or a State . . . in an administrative or judicially approved settlement may seek contribution from any person who is not party to [the] settlement.” So a litigant can bring a contribution claim once it has affirmed its potential li- ability in a settlement with the government. Third, each of § 9613(g)(3)’s triggers—a judgment and three kinds of settle- ments—formally recognizes a litigant’s actual or potential CERCLA liability.

These clues convince us that a contribution claim accrues when a litigant formally recognizes its CERCLA liability. Ap- plying that rule here, Cranbury Brick Yard’s contribution claim accrued in 2006, when Cranbury Brick Yard and NJDEP exe- cuted the amended Consent Order. By doing so, they recog- nized Cranbury Brick Yard’s potential liability, settled it, and acknowledged Cranbury Brick Yard’s “right to seek contribu- tion” from other polluters under § 9613(f)(3)(B). App. 170. 4. We decline to borrow the cost-recovery trigger. Cran- bury Brick Yard urges us not to borrow § 9613(g)(3)’s trigger, which is pegged to the date of accrual (here, the date of settle- ment). Instead, it invites us to borrow the trigger for cost- recovery actions: the date the cleanup begins. We decline this invitation. Cost recovery and contribution are two distinct causes of action authorized by two different statutory provisions. 42 U.S.C. §§ 9607(a), 9613(f). Likewise, their differing statutes of limitations are codified separately: one in a provision titled “Actions for recovery of costs,” the other in a provision titled “Contribution.” Id. § 9613(g)(2)–(3). Plus, Congress enacted the two causes of action six years apart. See Cooper Indus., 543 U.S. at 161–63 (canvassing the enact- ment of these causes of action). Given the differences between the two statutes of limitations, there is no textual basis to think that Congress intended to apply the trigger for cost-recovery actions to some contribution actions. See RSR Corp. v. Com- mercial Metals Co., 496 F.3d 552, 558 (6th Cir. 2007). In support of its argument, Cranbury Brick Yard cites two out-of-circuit cases that borrowed the cost-recovery trigger. See Appellant’s Br. 42–43 (citing Geraghty & Miller, Inc. v.

Conoco, Inc., 234 F.3d 917 (5th Cir. 2000) and Sun Co. v. Browning-Ferris, Inc., 124 F.3d 1187 (10th Cir. 1997)). Those courts viewed contribution claims as a subspecies of cost- recovery claims. See Geraghty, 234 F.3d at 924; Sun Co., 124 F.3d at 1191–92. But those cases are stale. More recent Supreme Court prec- edent has clarified that cost recovery and contribution are “clearly distinct” causes of action. Atl. Research, 551 U.S. at 138 (quoting Cooper Indus., 543 U.S. at 163 n.3). On that ba- sis, the Sixth Circuit held that Geraghty and Sun Co. “are no longer good law on this point.” Hobart Corp., 758 F.3d at 774. We agree. In sum, the trigger for the limitations period was the amended Consent Order, not the cleanup. We will affirm the District Court’s holding that “the statute of limitations began to run [on] February 16, 2006, when all parties to the 2006 Amendment, including NJDEP, signed it.” 2018 WL 4828410, at *6. B. Cranbury Brick Yard brought its contribution claim too late under either applicable statute of limitations

The clock started ticking when Cranbury Brick Yard joined the amended Consent Order. Now we must address the length of the limitations period. There are two possible options. Section 9613(g)(3) pro- vides a three-year limitations period for contribution claims. And under the Federal Tort Claims Act, “every civil action commenced against the United States shall be barred unless the complaint is filed within six years after the right of action first accrues.” 28 U.S.C. § 2401(a). But § 2401(a) is a backstop; a statutory scheme’s specific limitations period sometimes dis- places it. See Kannikal v. Att’y Gen. U.S., 776 F.3d 146, 150 (3d Cir. 2015); Mendoza v. Perez, 754 F.3d 1002, 1018 (D.C. Cir. 2014). Either way, Cranbury Brick Yard’s contribution claim is untimely. It accrued when Cranbury Brick Yard joined the amended Consent Order in 2006. But Cranbury Brick Yard did not sue until nine years later. Since Cranbury Brick Yard’s ac- tion would be untimely under either possible statute of limita- tions, we need not decide which statute applies. V. WE NEED NOT DECIDE IF CRANBURY BRICK YARD IS A BONA FIDE PROSPECTIVE PURCHASER

Cranbury Brick Yard also appeals the District Court’s hold- ing that it was not a bona fide prospective purchaser. This issue mattered only because the lack of that status might have ex- posed Cranbury Brick Yard to the government’s counterclaim for contribution, which it filed in response to Cranbury Brick Yard’s cost-recovery claim. But Cranbury Brick Yard has no cost-recovery claim, so the government’s counterclaim is now moot. Thus, bona-fide- prospective-purchaser status is irrelevant. And in any event, the amended Consent Order immunized Cranbury Brick Yard from contribution liability. So the District Court did not need to decide whether Cranbury Brick Yard was a bona fide pro- spective purchaser, and we need not reach this issue on appeal.

* * * * * By joining the amended Consent Order, Cranbury Brick Yard settled any possible liability to the State of New Jersey and thus extinguished its cost-recovery claim. While it still had a contribution claim, it brought that claim several years too late. The District Court properly granted summary judgment for the government on both claims, so we will affirm.