In this case, New York State Electric and Gas Corporation (“NYSEG”) sued FirstEnergy Corporation (“FirstEnergy”) under section 107(a) of the Comprehensive Environmental Response, Compensation, and Liability Act of 1980, 42 U.S.C. §§ 9601 et seq. (“CERCLA”), to recover certain costs incurred in remediating coal tar contamination at certain of NYSEG’s manufactured gas plants in upstate New York. NYSEG contends that FirstEnergy is hable as the successor to NYSEG’s former parent company, Associated Gas & Electric Company (“AGECO”), for a portion of the cleanup costs. FirstEnergy filed counterclaims against NYSEG and third-party claims against I.D. Booth, Inc. (“I.D. Booth”), the current owner of one of the sites, for cost contribution under section 113(f) of CERCLA
On July 11, 2011, following a bench trial, the United States District Court for the Northern District of New York (Peebles, M.J.) issued a decision and order holding that NYSEG was entitled to recover certain cleanup costs from FirstEnergy based on a veil-piercing theory, but limiting that recovery to certain sites. New York State Elec. & Gas Corp. v. FirstEnergy Corp., 808 F.Supp.2d 417, 499-502 (N.D.N.Y.2011) (“NYSEG”).1 The district court also found I.D. Booth liable for a portion of the cleanup costs at one site. Id. at 519.
We affirm in part, vacate in part, and remand.
STATEMENT OF THE CASE
A. The MGPs
This case arises from the cleanup of hazardous waste created at certain former manufactured gas plants (“MGPs”) in upstate New York, currently or formerly owned by NYSEG or its predecessor companies. MGPs began operating in the United States in the 1800s, producing gas used for cooking, lighting, and heating. The plants created gas by heating coal to very high temperatures in large ovens. The gas was then cleaned, processed and piped out for use. Unfortunately, as the gas cooled, it created a number of byproducts, including coal tar, which inevitably leaked from tar-handling equipment. Because coal tar is heavier than water, it tends to migrate in the subsurface, and travels underground from a site through the water table until it runs into a confining layer, such as bedrock. Coal tar also leaches into groundwater, causing groundwater contamination. The Environmental Protection Agency (“EPA”) listed coal tar as a hazardous waste in 1992. See 40 C.F.R. § 261.32(a) (2012); see also Identification and Listing of Hazardous Waste; CERCLA Hazardous Substance Designation; Reportable Quantity Adjustment; *218Coke By-Products Wastes, 57 Fed.Reg. 37,284-285 (Aug. 18, 1992) (codified at 40 C.F.R. Pts. 261, 271, & 302).
Most of the MGPs closed in the 1930s and 1940s when natural gas began to be delivered through interstate pipelines. In this case, all of the waste in dispute was manufactured before 1940. NYSEG or its predecessors owned at least thirty-eight MGP sites, including the sites at issue in this case.
B. The Cleanup
Most of the sites at issue were listed by the New York State Department of Environmental Conservation (“DEC”) in 1986 as Class “2a” sites on the Registry of Inactive Hazardous Waste Disposal Sites in New York.2 In the 1980s, NYSEG investigated all the MGP sites involved in this case, except Newark and Corning. In March 1994, DEC entered into a Consent Order with NYSEG addressing the investigation and cleanup of coal tar and associated hazardous substances at all the sites in this action, except Corning. Since then, DEC has retained oversight of the cleanup process and approved all cleanup projects at each MGP site at issue in this case.
C. Corporate History
The history of the corporations involved in this case is long and tortured. We relate only the points relevant to the issues before us.3
AGECO was incorporated as a public utility holding company in 1906. By 1907 it owned the common stock of several utility companies. Mergers of certain of its subsidiaries in 1916 and 1918 eventually led to what became known as the New York State Gas and Electric Corporation. In 1928, the latter entity changed its name to the New York State Electric Corporation, and a year later it adopted its current name, New York State Electric and Gas Corporation. Hence, NYSEG was created through the merger of certain AGECO subsidiaries.
Over the years, AGECO acquired other utility companies and MGPs, either directly or through other holding companies. In the 1930s, NYSEG acquired a number of MGPs from AGECO subsidiaries. By 1939, NYSEG had acquired all the MGPs at issue in this action from AGECO.
D. The Bankruptcy
On January 10, 1940, AGECO filed for bankruptcy. Pursuant to the reorganization plan, AGECO merged into AGE-CORP, which subsequently changed its name to General Public Utilities Corporation, which later became GPU. In 2001, GPU merged into FirstEnergy. Hence, FirstEnergy is the successor to AGECO.
On June 26,1945, during the bankruptcy proceedings, NYSEG’s board of directors adopted a resolution not to bring any claims against AGECO, instead assigning *219NYSEG’s claims to N.Y. PA NJ Utilities Company:
[T]hat in accordance with the request of N.Y. PA NJ Utilities Company dated May 9,1945, this Company shall take no action with respect to the filing of any claim or claims against the Estate of [AGECO] or the Estate of [AGECORPJ ...; provided, however, that in consideration therefor N.Y. PA NJ Utilities Company shall release this Corporation and its officers and directors from any liability arising from the omission of this Corporation to file such claim or claims and also from any liability for having made or approved allegedly excessive payments through various service corporations or funds prior to 1939; and provided, further that the Trustees of the above-mentioned Estates shall execute and deliver to this Corporation [an] appropriate covenant not to sue on account of any alleged failure to pay its pro rata share of any alleged Federal tax liability for the years 1927 to 1993, inclusive.
NYSEG, 808 F.Supp.2d at 433 (emphasis added).
E. Procedural History
This litigation began in April 2003, when NYSEG sued FirstEnergy under section 107(a) of CERCLA for cleanup costs at twenty-four MGPs in upstate New York. This number was reduced to seventeen sites before trial. During the trial, NY-SEG’s claims with respect to the Auburn Clark Street site were dismissed pursuant to Fed.R.Civ.P. 52(c), leaving sixteen sites at issue. NYSEG, 808 F.Supp.2d at 446 n. 11.
At trial, NYSEG alleged that it spent more than $94 million in cleanup costs on the sixteen sites through the end of 2009, and that it faced another $144 million in future cleanup costs. Id. at 428.
On July 11, 2011, the district court issued a decision and order, NYSEG, 808 F.Supp.2d 417, holding principally that:
1) NYSEG was not barred by a covenant not to sue from seeking contribution from FirstEnergy under CERCLA.
2) FirstEnergy was not liable for cleanup costs as an “owner.”
3) FirstEnergy was not directly liable for cleanup costs as an “operator.”
4) NYSEG was permitted to pierce the corporate veil to hold FirstEnergy hable as an “operator” for the period from 1922 to 1940.
5) NYSEG was not permitted to pierce the corporate veil for the period prior to 1922 or after 1940.
6) Cost recovery by NYSEG at the Norwich and Owego sites was precluded by the statute of limitations.
7) Cost recovery by NYSEG at the Plattsburgh site was not precluded by the statute of limitations.
8) A portion of a $20 million insurance payment to NYSEG could be used to offset costs assigned to FirstEnergy.
9) The court could rely on First-Energy’s coal tar production calculations for (1) the pre-1887 period and (2) the post-1930 period.
10) FirstEnergy was entitled to contribution from I.D. Booth based on I.D. Booth’s status as an owner of one of the sites.
An Amended Final Judgment was entered on September 7, 2011, awarding NY-SEG $29,715,225 for past and future cleanup costs for sixteen sites. I.D. Booth was held liable for $179,122 plus a share of future costs with respect to the Cortland-Homer site.
These appeals followed.
*220 DISCUSSION
A. CERCLA
Congress enacted CERCLA, 42 U.S.C. §§ 9601 et seq., to address the cleanup of hazardous waste by imposing strict liability for necessary cleanup costs incurred that are “consistent with the national contingency plan.” CERCLA § 107(a)(4)(B), 42 U.S.C. § 9607(a)(4)(B). Private parties who engage in cleanup activity can recover costs associated with such actions by bringing claims under either section 107(a) or section 113(f) of CERCLA against “potentially responsible parties” (“PRPs”). CERCLA created four classes of PRPs: (1) present owners and operators of facilities that accepted hazardous substances for transport; (2) past owners and operators of such facilities; (3) generators of hazardous substances; and (4) certain transporters of hazardous substances. 42 U.S.C. § 9607(a); see also Price Trucking Corp. v. Norampac Indus., Inc., 748 F.3d 75, 79-80 (2d Cir.2014).
Private parties have two options to recover their cleanup costs from other PRPs. First, under section 107(a), a property owner or operator who has spent money on cleaning up hazardous waste may seek reimbursement for cleanup costs from other PRPs. See Niagara Mohawk Power Corp. v. Chevron U.S.A., Inc., 596 F.3d 112, 120-21 (2d Cir.2010). This option, however, is limited to the “necessary costs of response ... consistent with the national contingency plan.” 42 U.S.C. § 9607(a)(4)(B).
Second, under section 113(f), “[a]ny person may seek contribution from any other person who is liable or potentially hable under [section 107(a) ] during or following any civil action under section 9606 of this title or under section 9607(a) of this title.” 42 U.S.C. § 9613(f)(1). In other words, a PRP who has been sued under section 107(a) to contribute to cleanup costs—even if it has not yet spent any money on cleanup activities—can seek contribution from other PRPs for cleanup costs, including from the initial plaintiff who sued the PRP under section 107(a). Here, NYSEG has sued FirstEnergy under section 107(a) and FirstEnergy has filed claims against NY-SEG and I.D. Booth under section 113(f).
We construe CERCLA liberally to advance its “dual goals of cleaning up hazardous waste and holding polluters responsible for their actions.” New York v. Next Millenium Realty, LLC, 732 F.3d 117, 124 (2d Cir.2013); see also Prisco v. A & D Carting Corp., 168 F.3d 593, 602 (2d Cir.1999) (“As a remedial statute, CERCLA should be construed liberally to give effect to its purposes.”) (quoting B.F. Goodrich v. Betkoski, 99 F.3d 505, 514 (2d Cir.1996), overruled on other grounds by New York v. Nat’l Serv. Indus., Inc., 352 F.3d 682, 685 (2d Cir.2003) (internal quotation marks omitted)); B.F. Goodrich Co. v. Murtha, 958 F.2d 1192, 1198 (2d Cir.1992) (“Because it is a remedial statute, CERCLA must be construed liberally to effectuate its two primary goals: (1) enabling the EPA to respond efficiently and expeditiously to toxic spills, and (2) holding those parties responsible for the releases liable for the costs of the cleanup.”).
B. Analysis
These appeals present the following issues: (1) whether NYSEG’s CERCLA claims against FirstEnergy are barred by the covenant not to sue; (2) whether AGE-CO is directly liable under CERCLA as an operator; (3) whether FirstEnergy is ha-ble to NYSEG on a piercing the corporate veil theory (a) based on AGECO’s control over NYSEG from 1922 to 1940, and (b) for contamination created by other AGE-CO subsidiaries before the subsidiaries were merged into NYSEG, and before *221AGECO owned the subsidiaries that were later merged into NYSEG; (4) whether NYSEG’s claims as to the Plattsburgh, Norwich, and Owego sites are time-barred; (5) whether the district court erred in calculating total gas production at the sites; (6) whether NYSEG’s recovery from First-Energy could be reduced by a portion of NYSEG’s $20 million insurance settlement; (7) whether NYSEG’s recovery should have been reduced either (a) to reflect the increased value of the remediat-ed properties or (b) because of NYSEG’s alleged delay in the remedial efforts; and (8) whether I.D. Booth is liable for a portion of cleanup costs.
We review the district court’s interpretation of the covenant not to sue, its decisions whether to pierce the corporate veil, and its statute-of-limitations determinations de novo. See Krumme v. West-Point Stevens Inc., 238 F.3d 133, 144 (2d Cir.2000) (waiver); Thrift Drug, Inc. v. Universal Prescription Adm’rs, 131 F.3d 95, 97 (2d Cir.1997) (veil piercing); Castagna v. Luceno, 744 F.3d 254, 256 (2d Cir.2014) (statute of limitations). We review the district court’s findings of fact for clear error and its conclusions of law de novo, Amalfitano v. Rosenberg, 533 F.3d 117, 123 (2d Cir.2008), and its evidentiary rulings for abuse of discretion, Boyce v. Soundview Tech. Grp., Inc., 464 F.3d 376, 385 (2d Cir.2006). We review the district court’s allocation of response costs for abuse of discretion. Goodrich Corp. v. Town of Middlebury, 311 F.3d 154, 168-69 (2d Cir.2002).
1. The Covenant Not To Sue
FirstEnergy argues the district court erred when it held that NYSEG’s claims were not foreclosed by the covenant not to sue that AGECO included in the resolution adopted by NYSEG’s board of directors during the bankruptcy proceedings in 1945. NYSEG, 808 F.Supp.2d at 503.
NYSEG argues that the covenant does not bar its claims because: (1) there is no evidence the covenant was ever executed and delivered and (2) the resolution is limited to claims of NYSEG in 1945 against the bankruptcy estates of AGECO and AGECORP in the bankruptcy proceedings.
A covenant not to sue is “an agreement by one having a present right of action against another not to sue to enforce such right.” Colton v. New York Hosp., 53 A.D.2d 588, 589, 385 N.Y.S.2d 65, 66 (1st Dep’t 1976); accord McMahan & Co. v. Bass, 250 A.D.2d 460, 461, 673 N.Y.S.2d 19, 21 (1st Dep’t 1998) (covenant not to sue. “constitutes an agreement to exercise forbearance from asserting any claim which either exists or which may accrue”). Although a covenant not to sue is distinct from a release, which is the “present abandonment of a right or claim,” Colton, 53 A.D.2d at 589, 385 N.Y.S.2d 65, a covenant not to sue has been “held to operate as a release ... in so far as the person to whom the covenant, in terms, runs, is concerned.” Shaw v. Crissey, 182 Misc. 27, 28, 43 N.Y.S.2d 237, 239 (Sup.Ct. Saratoga Cnty.1943).
Here, however, the covenant never became operative because a condition precedent to its validity never occurred. Under New York Law, “when there is a ‘condition precedent to the formation or existence of the contract itself ... no contract arises ‘unless the condition occurs.’ ” Adams v. Suozzi, 433 F.3d 220, 227 (2d Cir.2005) (quoting Oppenheimer & Co., Inc. v. Oppenheim, Appel, Dixon, & Co., 86 N.Y.2d 685, 690, 636 N.Y.S.2d 734, 660 N.E.2d 415 (1995)) (emphasis omitted). The board resolution explicitly stated that NYSEG releases its claims “provided ... that [AGECO Trustees] shall execute and *222deliver to [NYSEG] an appropriate covenant not to sue” for certain alleged failures. NYSEG, 808 F.Supp.2d at 433. Yet FirstEnergy failed to offer proof that satisfied this condition. Because First-Energy failed to carry its burden of proof to show that the conditions precedent to the existence of the NYSEG covenant occurred, we affirm the district court’s ruling that the covenant does not bar NYSEG’s claims.4
2. AGECO’s direct liability as a CERCLA “operator”
NYSEG argues that FirstEnergy is directly liable as an operator of the Cortland-Homer, Ithaca-Court Street, and Norwich MGPs between 1906 and 1910 and the Oneonta MGP between 1916 and 1922 because AGECO “directly managed and operated” these MGPs. NYSEG Br. at 24.
Under CERCLA, “any person who operates a polluting facility is directly liable for the costs of cleaning up the pollution ... regardless of whether that person is the facility’s owner [or] the owner’s parent corporation.” United States v. Bestfoods, 524 U.S. 51, 65, 118 S.Ct. 1876, 141 L.Ed.2d 43 (1998) (internal citations omitted). An “operator” is defined as “someone who directs the workings of, manages, or conducts the affairs of a facility.” Id. at 66, 118 S.Ct. 1876. To be held liable under CERCLA, an operator must “manage, direct, or conduct operations specifically related to pollution, that is, operations having to do with the leakage or disposal of hazardous waste, or decisions about compliance with environmental regulations.” Id. at 66-67, 118 S.Ct. 1876. Whether an entity directly operates a facility is a factual issue that we review for clear error. See Am. Cyanamid Co. v. Capuano, 381 F.3d 6, 23 (1st Cir.2004).
There are three circumstances under which a parent can be held liable as a direct operator of a subsidiary’s facilities: (1) “when the parent operates the facility in the stead of its subsidiary or alongside the subsidiary in some sort of a joint venture”; (2) when “a dual officer or director ... departfs] so far from the norms of parental influence exercised through dual officeholding as to serve the parent, even when ostensibly acting on behalf of the subsidiary”; and (3) when “an agent of the parent with no hat to wear but the parent’s hat ... manage[s] or direct[s] activities at the facility.” Bestfoods, 524 U.S. at 71, 118 S.Ct. 1876. In considering these questions, a court must focus on the relationship between the parent and the facility in question, not the parent and the subsidiary. See id. at 67-68, 118 S.Ct. 1876.5
Here, the district court found that AGE-CO was not an operator during this period because it did not sufficiently “participate[] in the activities” of the facilities. Id. at 68, 118 S.Ct. 1876. Specifically, the district court relied on the fact that:
[E]ach MGP facility retained its own superintendent on site who was responsible for the day-to-day activities, and *223there is no evidence that the superintendent reported to and was controlled by AGECO and the service companies, as distinct from the corporate management and board of directors of NYSEG and those of other subsidiary utility companies.
NYSEG, 808 F.Supp.2d at 493.
The district court noted that while NYSEG and AGECO entered into service agreements under which AGECO was “retained to provide such services as general operator and financial manager of NY-SEG’s properties,” those agreements did not “reveal the type of management and control over polluting activities envisioned by Bestfoods as necessary to support a finding of direct operator liability.” Id. at 493-94. The fact that AGECO took steps to “monitor and control” its subsidiaries’ activities, including arranging for these service agreements, was consistent with AGECO’s role as a parent corporation managing the activities of its subsidiary. Id. As the Bestfoods court observed:
Activities that involve the facility but which are consistent with the parent’s investor status, such as monitoring of the subsidiary’s performance, supervision of the subsidiary’s finance and capital budget decisions, and articulation of general policies and procedures, should not give rise to direct liability.
524 U.S. at 72, 118 S.Ct. 1876 (alterations omitted).
NYSEG’s arguments to the contrary are not persuasive because the activities it cites are consistent with acceptable practices of a parent corporation. For example, NYSEG points to a series of reports issued by the Federal Trade Commission between 1932 and 1934, which stated that after AGECO’s owners transferred the stock of Ithaca Gas Light, Norwich Gas Company, and Homer & Cortland Gas Light Company (the owners of the Cortland-Homer, Ithaca-Court Street, and Norwich MGPs) to AGECO, AGECO brought the companies under “common control and management” and managed and operated their MGPs from 1906 to 1910. NYSEG Br. at 5. More specifically, NYSEG cites Brown’s Directory of American Gas Companies (“Brown’s Directory”), which reported that the same AGE-CO personnel held the offices of President, Vice-President, Secretary, and General Manager of the three MGP’s during this period. NYSEG also points to the previously mentioned service contracts with W.S. Barstow & Company and J.G. White & Company, which were in effect at various points between 1910 and 1922 as evidence of direct control of the Cortland-Homer, Ithaca-Court Street, Norwich, and Oneonta MGPs.
The district court did not err in holding that these general allegations failed to support a finding of direct operator liability as they do not show that AGECO operated the facility “in the stead of its subsidiary.” See Bestfoods, 524 U.S. at 71, 118 S.Ct. 1876. The holding of dual officerships and directorships, without more, is insufficient to establish operator liability. Id. at 70-71, 118 S.Ct. 1876. To warrant direct liability, a dual officer or director must “depart so far from the norms of parental influence exercised through dual office-holding as to serve the parent, even when ostensibly acting on behalf of the subsidiary in operating the facility.” Id. The district court did not commit clear error in finding no such departure in this case. F.D.I.C. v. Providence Coll., 115 F.3d 136, 140 (2d Cir.1997) (“We will set aside a district court’s findings of fact following a bench trial only if those findings are clearly erroneous.”). Similarly, the supervisory activities engaged in by AGECO (monitoring performance, supervising finances, etc.) fall within the parameters of the par*224ent-subsidiary relationship. Bestfoods, 524 U.S. at 72, 118 S.Ct. 1876. Thus, we affirm the district court’s decision not to find direct operator liability.
3. Piercing the Corporate Veil
а. Contamination Created by NYSEG.
FirstEnergy argues that the district court erred in piercing the corporate veil to hold it liable for cleanup costs for pollution created by NYSEG.
It is fundamental that a parent is considered a legally separate entity from its subsidiary, and cannot be held liable for the subsidiary’s actions based solely on its ownership of a controlling interest in the subsidiary. Bestfoods, 524 U.S. at 61, 118 S.Ct. 1876; Carte Blanche (Singapore) Pte., Ltd. v. Diners Club Intern., Inc., 2 F.3d 24, 26 (2d Cir.1993). Under New York law,6 however, a parent can be held liable for the actions of a subsidiary where a plaintiff shows: (1) the parent corporation dominates the subsidiary in such a way as to make it a “mere instrumentality” of the parent; (2) the parent company exploits its control to “commit fraud or other wrong”; and (3) the plaintiff suffers an unjust loss or injury as a result of the fraud or wrong. Wm. Passalacqua Builders, Inc. v. Resnick Developers S., Inc., 933 F.2d 131, 138 (2d Cir.1991) (internal quotations omitted); accord Bestfoods, 524 U.S. at 63-64, 118 S.Ct. 1876.
As to the first prong, courts consider the following factors to determine the degree of domination by the parent:
(1) the absence of the formalities and paraphernalia that are part and parcel of the corporate existence, i.e., issuance of stock, election of directors, keeping of corporate records and the like[;]
(2) inadequate capitalization^]
(3) whether funds are put in and taken out of the corporation for personal rather than corporate purposes[;]
(4) overlap in ownership, officers, directors, and personnel^]
(5) common office space, address and telephone numbers of corporate entities^]
(6) the amount of business discretion displayed by the allegedly dominated corporation^]
(7) whether the related corporations deal with the dominated corporation at arms length[;]
(8) whether the corporations are treated as independent profit centers[;]
(9) the payment or guarantee of debts of the dominated corporation by other corporations in the group[;] and
(10) whether the corporation in question had property that was used by other of the corporations as if it were its own.
Passalacqua, 933 F.2d at 139. The ultimate question is whether “the policy behind the presumption of corporate independence and limited shareholder liability—encouragement of business development—is outweighed by the policy justifying disregarding the corporate form—the need to protect those who deal with the corporation.” Id.
NYSEG sought to pierce the corporate veil here in three respects: (a) AGECO’s control over NYSEG from 1922 to 1940; (b) contamination created by other subsidiaries before they were merged into NY-SEG; and (c) contamination created by other subsidiaries before AGECO owned *225the subsidiaries that were merged into NYSEG.
FirstEnergy argues that the district court failed to make sufficient NY-SEG-specifie findings to warrant piercing the corporate veil.7 We disagree. The district court made numerous findings regarding AGECO’s domination of NYSEG when it carefully considered the Passalac-qua factors in reaching its decision. These include findings specific to NYSEG as well as more general findings regarding AGE-CO’s abuse of all its subsidiaries, including NYSEG. While it is true that not every factor is present in this case, that is not a requirement under Passalacqua, which recognized the “infinite variety of situations that might warrant disregarding the corporate form.” Passalacqua, 933 F.2d at 139; see also Freeman v. Complex Computing Co., Inc., 119 F.3d 1044, 1053 (2d Cir.1997) (holding “[n]o one [Passalacqua] factor is decisive”).
As an initial matter, the district court found that between 1922 and 1940 AGECO was dominated and controlled by Howard C. Hopson and John I. Mange. NYSEG, 808 F.Supp.2d at 436. Hopson and Mange had acquired all AGECO’s shares of voting stock by April 1923. Id. One way in which Hopson controlled the subsidiary companies was by holding the directors’ undated, signed resignations in hand. Id. at 499.
While the district court found that NY-SEG was not undercapitalized, it did find that Hopson freely transferred funds in and out of AGECO and its subsidiaries, and that the subsidiaries were considered “mere pockets” of AGECO. Id. For example, the New York Public Service Commission (“PSC”) examined the books, records, accounting methods and documents of NYSEG, as well as other subsidiaries, from 1934 to 1938. Id. at 439-40. It issued a report (the “Brewster Report”) finding that AGECO and its affiliates “siphon[ed]” off funds from the operating companies and deposited them into the “pockets of those individuals and corporations engaged in milking the operating companies through the device of servicing and management contracts.” Id. at 439 (internal quotation marks omitted). It also found that Hopson and Howard C. Hopson & Company defrayed personal expenses with funds from the operating companies within the AGECO system. Id. at 440.
Furthermore, the district court found abuses by the Utility Management Corporation (“UMC”), which charged operating companies in the AGECO system, including NYSEG, a management fee of 2.5 percent of their gross revenues even though no management employees from UMC were located on any property owned by the operating companies. Id. The court also found substantial overlap in officers, *226directors, and personnel of AGECO and the subsidiaries. For example, during virtually all the years from 1922 to 1940, Mange served as president of AGECO and a director of NYSEG, and Hopson was on the NYSEG board from 1927 to 1934. Id. at 442-43.
Between 1922 and 1940, board meetings for NYSEG and various other AGECO operating companies were usually held in New York City at or near AGECO’s office at 61 Broadway, which was also the primary location of Hopson’s accounting and financial organization that rendered financial, legal, accounting, and auditing services to AGECO and its subsidiaries. Id. at 442.
The district court found that the terms of the service contracts between the service companies run by AGECO and the operating utility companies, including NY-SEG, left the utility companies with “no vestige of independent authority or control” because “[u]nder the provisions of these contracts, the service corporations manage, dominate, and practically operate the utilities.” Id. at 439 (quoting 1932 PSC Report).
The district court found that AGECO and the subsidiaries did not deal at arms length, as no one represented NYSEG or any of the other subsidiaries in the service contract negotiations. Id. at 443. In addition, AGECO loaned money to NYSEG and guaranteed debts by others to NY-SEG. Id. at 499. Considering the totality of these NYSEG-specific findings, we conclude the district court correctly found that veil piercing was warranted. See Passalacqua, 933 F.2d at 139.8
The question remains as to when AGE-CO’s domination of NYSEG ended. First-Energy argues that any domination ended on January 10,1940, the day the bankruptcy was filed. The district court held that AGECO’s domination did not end until December 31, 1940. NYSEG, 808 F.Supp.2d at 539.
The district court’s conclusion that domination continued throughout 1940 is not supported by the record.9 As the party seeking to pierce the corporate veil, NY-SEG has the “heavy burden” of demon*227strating domination, TNS Holdings v. MKI Secs. Corp., 92 N.Y.2d 335, 339, 680 N.Y.S.2d 891, 703 N.E.2d 749 (1998), yet it adduced a paucity of affirmative evidence that AGECO continued to control NYSEG after filing for bankruptcy. Indeed, in its findings of fact the district court found that “NYSEG ... failed to prove a basis for veil-piercing ... for the period following AGECO’s bankruptcy filing.” NYSEG, 808 F.Supp.2d at 502 (emphasis added). It was only later, in its conclusions of law, that the district court concluded that liability would extend throughout 1940 because there was testimony from NYSEG’s expert that “control and dominance of AGECO over its subsidiaries extended well into 1941.” Id. at 539. We conclude that these seemingly contradictory statements betray the lack of affirmative evidence needed for NYSEG to have carried its burden of proving dominance and control sufficient to warrant piercing the corporate veil past AGECO’s bankruptcy filing.
b. Contamination Created by Other AGECO Subsidiaries.
The parties have raised two additional issues regarding veil piercing: (1) whether the district court erred in piercing the corporate veil to hold FirstEnergy liable for contamination created by other AGE-CO subsidiaries before those subsidiaries were merged into NYSEG by AGECO; and (2) whether the district court erred in refusing to pierce the corporate veil to hold FirstEnergy liable for contamination that was created by companies before they were owned by either NYSEG or AGECO.
Between 1929 and 1939, AGECO merged five subsidiaries into NYSEG: (1) Eastern New York Electric and Gas Company (owner of the Granville, Mechanic-ville and Plattsburgh MGPs); (2) Elmira Light, Heat & Power Corporation (owner of the Elmira MGP); (3) New York Central Electric Corporation (owner of the Corning, Dansville, Newark, and Penn Yan MGPs); (4) Empire Gas & Electric Company (owner of the Geneva MGP); and (5) Owego Gas Corporation (owner of the Ow-ego MGP).10 NYSEG, 808 F.Supp.2d at 501. NYSEG also acquired the Federal-New York Company, Inc., owner of the Goshen MGP during this period. Id. Hence, these five subsidiaries owned ten MGPs.
According to NYSEG, FirstEnergy absorbed all of the subsidiaries’ preexisting liabilities when NYSEG’s veil was pierced, even those liabilities that were incurred before either NYSEG or FirstEnergy owned the subsidiaries. NYSEG argues that because AGECO merged the five subsidiaries into NYSEG, these subsidiaries fall into one of the exceptions to the general rule regarding liabilities acquired through asset purchases. We disagree.
When a corporation purchases the assets of another corporation, it does not acquire its liabilities, unless one of four exceptions applies. See New York v. Nat’l Serv. Indus., Inc., 460 F.3d 201, 209 (2d Cir.2006). The four exceptions are: (1) the buyer “expressly or impliedly assumed the predecessor’s tort liability, (2) there was a consolidation or merger of seller and purchaser, (3) the purchasing corporation was a mere continuation of the selling corporation, or (4) the transaction [wa]s entered into fraudulently to escape such *228obligations.” Id. (internal quotation marks omitted). Here, because there is no question that AGECO merged these five subsidiaries into NYSEG, NYSEG acquired their liabilities, including liabilities for past contamination. See Maline v. City of Utica, 267 A.D.2d 1022, 1022, 701 N.Y.S.2d 202, 203 (4th Dep’t 1999) (as a result of merger with co-defendant, defendant became the owner of co-defendant’s property and assumed all of its “liabilities, obligations and penalties”). That does not answer the question, however, as to what extent those liabilities flowed to NYSEG’s parent, AGECO, under a veil-piercing theory.
1. Pre-Merger Liability
FirstEnergy argues that the district court erred in piercing the corporate veil to hold it liable for contamination created at the ten MGPs owned by AGECO subsidiaries before the subsidiaries were merged into NYSEG. FirstEnergy notes that six of the AGECO-owned MGPs ceased producing gas before AGECO merged these subsidiaries with NYSEG.11
As a general matter, under a veil-piercing theory, a parent can only be liable for a wrong committed by the subsidiary under the influence of the parent. Passalacqua, 933 F.2d at 138. The use of control to commit a wrong is a key component to veil piercing. Id. (control of the subsidiaries by the parent “is the key; the control must be used to commit a fraud or other wrong that causes plaintiffs loss”); Bedford Affiliates v. Sills, 156 F.3d 416, 431 (2d Cir.1998) (court must find that parental control was used to cause a wrong), overruled on other grounds by W.R. Grace & Co.-Conn. v. Zotos Intern., Inc., 559 F.3d 85, 89-90 (2d Cir.2009). Domination and control by themselves are not sufficient to support a finding of veil piercing. Id. (“Mere domination or control of the corporation is insufficient to permit a court to disregard the existence of the corporate entity.”).
The district court held that the veil of the non-NYSEG subsidiaries could be pierced to hold FirstEnergy liable for contamination created while the subsidiaries were dominated by AGECO but before they were merged into NYSEG. NYSEG, 808 F.Supp.2d at 491-500. (“[E]quity requires piercing of the NYSEG corporate veil as well as the other operating companies within the AGECO Empire and ultimately folded into NYSEG ... during the times between 1922 and 1940 when AGE-CO acquired domination and control over the various subsidiary operating utility companies owning those facilities.”). The district court based this holding primarily on the testimony of NYSEG’s expert, Jonathan Macey, who stated that AGECO treated these other operating companies in the same manner as NYSEG. Id. at 443.
The district court, however, did not make any specific veil-piercing findings or discuss the Passalacqua factors with respect to AGECO’s alleged control over the non-NYSEG subsidiaries before NY-SEG acquired them. The district court did not make any findings concerning, for example, capitalization, director/officer ov*229erlap, and observance of corporate formalities, with respect to these subsidiaries during this time period. Hence, the district court’s findings are insufficient to support a conclusion of veil piercing in this respect. See William Wrigley Jr. Co. v. Waters, 890 F.2d 594, 601-02 (2d Cir.1989) (vacating finding of veil piercing where evidence showed individual defendant was personally involved with corporation’s work but not that he controlled corporation and was using it for fraudulent ends). Accordingly, we conclude that the district court erred by piercing the corporate veil of the five subsidiaries to hold FirstEnergy liable for contamination that occurred while under AGECO’s control, but before the subsidiaries were merged into NYSEG.12
2. Pre-Ownership Liability
NYSEG then goes one step further with its argument. Because all five of these subsidiaries were merged into NYSEG during the period when AGECO improperly dominated NYSEG (between 1922 and 1940), NYSEG argues that AGECO is responsible for all of the subsidiaries’ liabilities, even those that pre-date AGECO’s ownership. NYSEG reasons that all of the subsidiaries’ liabilities became NY-SEG’s liabilities as a result of the merger, even those incurred before the subsidiaries were purchased by AGECO. Thus, once NYSEG’s veil was pierced, all of its liabilities—including those inherited from the merged subsidiaries—became AGECO’s liabilities.
Courts will pierce the corporate veil to prevent fraud or achieve equity by imposing a corporate obligation upon a parent. Matter of Morris v. New York State Dept. of Taxation and Fin., 82 N.Y.2d 135, 141, 603 N.Y.S.2d 807, 623 N.E.2d 1157 (N.Y.1993). But to successfully pierce the veil, a plaintiff must show both domination of the corporation and that “such domination was used to commit a fraud or wrong against the plaintiff which resulted in plaintiffs injury.” Id.
Here, the “fraud” or “wrong” committed by AGECO would be the merger of the subsidiaries into NYSEG, not the creation of coal tar. There is nothing in the record to suggest AGECO was directing the creation of coal tar at the subsidiaries prior to purchasing them. While AGECO may have sought to merge its subsidiaries into NYSEG to further its financial improprieties, NYSEG does not allege AGECO did so to avoid CERCLA liability. Environmental liability for spilling coal tar was not a major concern at the time these subsidiaries were merged. AGECO’s domination was not used to commit a fraud or wrong against NYSEG-re-garding pollution, and domination by itself is not enough to justify veil piercing—it must be accompanied by a showing of wrongful or unjust action toward the plaintiff. Id. at 142, 603 N.Y.S.2d 807, 623 N.E.2d 1157 (plaintiff must show that owners “abused the privilege of doing business in the corporate form to perpetrate a wrong or injustice against that party such that a court in equity will intervene”); see also Freeman v. Complex Computing Co., 119 F.3d 1044, 1053 (2d Cir.1997); Passalacqua, 933 F.2d at 138 (control by parent must be used “to commit a fraud or other wrong that causes plaintiffs loss”). Thus, the district court correctly declined to pierce the corporate veil to hold AGECO responsible for contamination on sites that occurred prior to when either NYSEG or *230AGECO owned the sites. NYSEG, 808 F.Supp.2d at 500.
4. Statute of Limitations Regarding the Plattsburgh, Norwich, and Owego MGP Sites
The district court held that the statute of limitations barred NYSEG from recovering cleanup costs from FirstEnergy for the Norwich and Owego sites, but not the Plattsburgh site. NYSEG, 808 F.Supp.2d at 504-14. FirstEnergy argues that cost-recovery actions for cleanup expenses at the Plattsburgh, Norwich, and Owego sites are all barred by CERCLA’s six-year statute of limitations on remedial actions. See 42 U.S.C. § 9613(g)(2)(B). Because we conclude that the cleanup at the Norwich and Owego sites were remedial actions and the cleanup at the Plattsburgh site was a removal action, we affirm the district court’s decisions.
Removal and remedial actions are governed by different statutes of limitations. For removal actions, a party must seek to recoup cleanup costs within three years “after completion of the removal action.” Id. at § 9613(g)(2)(A). For remedial actions, a party must seek to recoup costs within six years “after initiation of physical on-site construction of the remedial action.” Id. at § 9613(g)(2)(B).13 Because a statute of limitations is an affirmative defense, FirstEnergy bears the burden of proof to show it bars the claims. See Overall v. Estate of Klotz, 52 F.3d 398, 403 (2d Cir.1995); Yankee Gas Servs. Co. v. UGI Utils, Inc., 616 F.Supp.2d 228, 269 (D.Conn.2009).
Whether a suit to recover response costs under section 107 of CERCLA is a “removal action” or a “remedial action” is a question of law that we review de novo. Next Millenium, 732 F.3d at 126; United States v. W.R. Grace & Co., 429 F.3d 1224, 1234 (9th Cir.2005) (‘Whether the ... cleanup activity was a removal action — or, on the other hand, a remedial action in removal action’s clothing — is a question of statutory interpretation.”).
Removal actions are generally clean-up measures taken in response to immediate threats to public health and safety. See 42 U.S.C. § 9601(23);14 see also Next Millenium, 732 F.3d at 124-25; W.R. Grace & Co., 429 F.3d at 1244 (“Courts have ... stressed the immediacy of a threat in deciding whether a cleanup is a removal action.” (collecting cases)); Minnesota v. Kalman W. Abrams Metals, Inc., 155 F.3d 1019, 1024 (8th Cir.1998) (removal actions are those “taken to counter imminent and *231substantial threats to public health and welfare”). Removal actions are also generally designed to address contamination at its endpoint and not to permanently remediate the problem. See 42 U.S.C. § 9601(24); Next Millenium, 732 F.3d at 127; W.R. Grace & Co., 429 F.3d at 1247 (finding cleanup activity to be removal action where it did not “fully eliminate the public health threat or amount to a full-blown remediation”).
Remedial actions are typically actions designed to permanently remediate hazardous waste. 42 U.S.C. § 9601(24);15 see also Next Millenium, 732 F.3d at 125; Schaefer v. Town of Victor, 457 F.3d 188, 195 (2d Cir.2006) (remedial actions are “generally long-term or permanent containment or disposal programs” (quoting New York v. Shore Realty Corp., 759 F.2d 1032, 1040 (2d Cir.1985) (internal quotation marks omitted))); California ex rel. Cal. Dep’t of Toxic Substances Control v. Neville Chem. Co., 358 F.3d 661, 667 (9th Cir.2004) (“remedial actions generally are permanent responses” (quoting Geraghty & Miller, Inc. v. Conoco, Inc., 234 F.3d 917, 926 (5th Cir.2000) (internal quotation marks omitted))).
a. Plattsburgh
The district court held that NYSEG’s claim to recover costs for the cleanup at the Plattsburgh MGP pursuant to a 1994 DEC Consent Order was timely. NYSEG, 808 F.Supp.2d at 510-11. FirstEnergy argues that NYSEG’s claim was untimely because the cleanup NYSEG conducted in 1981 was a remedial action, triggering the six-year statute of limitations. Although NYSEG is not seeking to recover cleanup costs associated with the 1981 Consent Order, id. at 511 n. 53, FirstEnergy contends that the 1981 cleanup triggered the statute of limitations.16
The Plattsburgh MGP, which sits adjacent to the south bank of the Saranac River, functioned as a coal gasification plant from 1896 to 1960. Coal tar had been kept on site in unlined ponds since 1898. As coal tar is heavier than water, it migrated from these ponds through the subsurface soils into the Saranac River, creating both soil and groundwater contamination.
*232In 1975, NYSEG began investigating the coal tar seeping into the Saranac River. Id. at 482. In mid-1980, DEC became aware that coal tar was seeping into the river. In 1981, NYSEG entered into a consent order with DEC to undertake work to prevent coal tar from reaching the Saranac River from a tar lagoon located across the street from the MGP. Id. This work was also to include cleaning up coal tar that had already migrated from the lagoon into the Saranac River, roughly thirty feet away. Id. at 509. As part of this effort, NYSEG built a slurry wall around the lagoon to isolate the coal tar, as well as a second slurry wall to catch any coal tar that made it past the first wall. It also excavated contamination from the river, filled in excavated areas with clean fill, capped containment areas, and constructed a treatment system for the water collected from behind the slurry wall. Id. This work did not involve the area where the former MGP plant was located. Id. at 482. This project took one year and cost less than $2 million.
Unfortunately, by the 1990s, coal tar was found in the river again. Between October 1997 and November 2000, NYSEG conducted further study of the Plattsburgh site. Id. at 483. Between April and August 2002, NYSEG excavated and removed waste from three gas holder foundations, coal-tar-containing process pipe, and other MGP associated structures. DEC approved this work in 2008. Id.
In 2004, DEC issued a Record of Decision (“ROD”) requiring additional remedial efforts. Id. This work included excavation of the former MGP tar lagoon and surrounding areas where coal tar had migrated to the subsurface, as well as excavation of the contaminated sediment from the parts of the Saranac River immediately adjacent to the MGP site. Id. The excavated soils were disposed of off-site. Id. NYSEG performed the ROD-mandated work between 2006 and 2009. Id. This included paying the City of Plattsburgh approximately $900,000 to move a substation and associated electrical lines owned by the city’s municipal power company from the northeast portion of the site. Id. NYSEG also relocated a twenty-one inch sewer crossing, constructed a stabilized soil barrier wall, and removed 150,816 tons of soil “down to till or bedrock inside the wall.” Id.
In 2009, NYSEG prepared another report which proposed “dewater[ing] the Saranac River channel” to permit further excavation of coal-tar-containing sediment. Id. NYSEG has implemented the first phase of this project and was scheduled to begin the second phase in 2011. Id. at 483-84. Between 1994 and 2009, NYSEG incurred $29,086,329.86 in response costs. Id. at 484.
FirstEnergy argues that because the cleanup pursuant to the 1981 Consent Order was a remedial action that triggered the six-year statute of limitations, all later cost recovery efforts at the site are time-barred, including the work done pursuant to the 1994 Consent Order. The district court disagreed, and held instead that the statute of limitations was not triggered because at the time the 1981 cleanup was conducted, coal tar was not a hazardous substance:
The work performed at Plattsburgh under the 1981 Consent Order ... cannot properly be regarded as consistent with the EPA’s remedial plan under CERC-LA based upon more modern notions regarding the hazards presented by the presence of coal tar and soils contaminated with coal tar, and thus the actions taken at Plattsburgh did not commence the running of the statute of limitations.
Id. at 510.
The district court also noted that the cleanup was “far more akin to a removal *233action” than a remedial action because it only addressed work “performed at a discrete portion of the site, which was significantly removed from the location of the former MGP facility [the source of contamination],” and was designed to contain the contamination migrating from the lagoon to the Saranac River. Id. Accordingly, the district court concluded that the 1981 removal action did not trigger the statute of limitations for the later remedial effort for which NYSEG now seeks contribution.
Assuming the 1981 cleanup was a triggering event, we agree with the district court that the 1981 cleanup was more akin to a removal than a remedial action. This cleanup was a discrete project designed both to prevent coal tar from reaching the river and to remove it from the river. See NYSEG, 808 F.Supp.2d at 509. It was not designed to clean up contamination at the source, namely by removing the coal tar from the area around the MGP. Id. at 482. Instead, as the district court pointed out, “[t]he evidence at trial reflected that neither DEC nor NYSEG envisioned the project as encompassing a CERCLA-quality clean-up of the entire MGP site.” Id. at 538 (emphasis added). As we noted in Next Millenium, measures taken to minimize and mitigate contamination, but not to permanently eliminate it, are properly classified as removal actions. 732 F.3d at 127; see also W.R. Grace & Co., 429 F.3d at 1247 (cleanup activity was a removal action where it did not “fully eliminate the public health threat or amount to a full-blown remediation”).
Here, the 1981 cleanup was an immediate response to a health concern about coal tar in the Saranac River. DEC first became aware of the problem in the mid-1980s and by 1981 had a Consent Order in place with NYSEG. Moreover, the cleanup was not a “permanent” solution designed to remove the ultimate source of contamination at the MGP. Rather, it involved building slurry walls to contain coal tar from further migrating into the river. These actions were not “designed to remedy the underlying source of the contamination, namely, the hazardous waste at the [MGP].” Next Millenium, 732 F.3d at 127; accord Schaefer, 457 F.3d at 195.
The work in 2002 and pursuant to the 2004 ROD, in contrast, was designed to remediate the pollution at its source. It included excavating the former MGP tar lagoon and surrounding area, moving an entire substation and corresponding electrical wires, the relocation of a sewer line, and the removal of over 150,000 tons of soil. The 2004 remedial action was clearly designed to permanently remediate the coal tar on the site.
In addition, the 2004 cleanup has already taken several years and cost over $29 million, with costs expected to rise to $54 million. This is in keeping with a remedial, rather than removal, action. See 42 U.S.C. § 9604(c)(1) (generally removal measures “shall not continue after $2,000,000 has been obligated for response actions or 12 months has elapsed from the date of initial response”); accord 40 C.F.R. § 300.415(b)(5). In sum, the district court correctly held that, because the 1981 cleanup was a removal action, it did not trigger the statute of limitations for the later remedial actions.17
*234 b. Norwich
The district court held that NYSEG’s cost recovery claims for its 1997 cleanup at Norwich were barred by the six-year statute of limitations on remedial actions. NYSEG argues that the 1997 cleanup was a removal action, subject to a different statute of limitations.
The Norwich MGP produced gas from 1863 to 1952. In 1993, NYSEG began a three-phase Interim Remedial Measure (“IRM”) in conjunction with DEC to clean up this site. In Phase I, which ran from 1993 to the last quarter of 1994, NYSEG excavated the former distribution holder area and stockpiled the soil. In Phase II, which ran from May to September 1996, NYSEG transported and thermally treated 1,600 tons of the stockpiled soil.
In Phase III, which ran from April to August 1997, NYSEG removed two or more feet of soil from the surface of the entire site, as well as the former relief holder, the former tar well, and process piping. This project was conducted to protect potential customers at an adjacent supermarket to be constructed shortly thereafter. In total, more than 11,000 tons of soil were removed. Phase III cost less than $2 million and was completed in four months. This is the cleanup for which NYSEG is seeking to recover costs.
FirstEnergy argues that the three phases of the IRM constituted one remedial project that commenced in 1997, and thus NYSEG’s cost recovery action is barred by the six-year statute of limitations. NY-SEG argues its claims are timely because all three phases of the IRM were removal actions, including the 1997 IRM. According to NYSEG, the 1997 IRM was a discrete project undertaken to address an immediate health hazard to future customers at a soon-to-be-built nearby supermarket. NY-SEG also argues that because DEC approved the work as an IRM, it is a removal action.
The district court correctly found that the three-phase IRM cleanup was one remedial cleanup. First, NYSEG stipulated that the Norwich IRM was a single action comprised of three phases. Second, the cleanup was not designed to address an imminent health concern. See Next Millenium, 732 F.3d at 124-25 (holding that removal actions are cleanup measures taken in response to immediate threats to public health and safety). As the court noted, “immediate action ... was not required to address high levels of toxic chemicals.” NYSEG, 808 F.Supp.2d at 513. Similarly, in its 1987 preliminary assessment, EPA deemed cleanup at the site a “medium priority,” rather than a high priority. Id. Hence, this was not an immediate threat to public health and safety.
Third, while IRMs can resemble removal actions, they can also resemble remedial actions. Under New York law, IRMs are defined as:
[Ajctivities to address both emergency and non-emergency site conditions, which can be undertaken without extensive investigation and evaluation, to prevent, mitigate or remedy environmental damage or the consequences of environmental damage attributable to a site.
N.Y. Comp.Codes R. & Regs. tit. 6, § 375-1.2(ab). This definition encompasses elements of both removal (emergency) and remedial (non-emergency) actions, as well as elements common to both (preventing or mitigating environmental damage). See 42 U.S.C. § 9601(23) (defining “removal” to include actions to “prevent, minimize, or mitigate damage to the public health or welfare or to the environment”); id. § 9601(24) (defining “remedial” to include actions to “prevent or minimize the release of hazardous substances” so that they do *235not endanger public health or the environment).
DEC’S present internal guidance and the guidance in effect during the Norwich IRM make clear that IRMs are not always removal actions. See, e.g., New York State Department of Environmental Conservation, DER-10/Technieal Guidance for Site Investigation and Remediation, at 35 (May 3, 2010), available at http://www.dec. ny.gov/docs/remediation_hudson_pdf/der 10.pdf (describing “[a]n emergency IRM” as well as “[a] non-emergency or non-time critical IRM ... which may be undertaken at any time during the course of the remedial program”); Pi’s. Exh. 4, New York State Department of Environmental Conservation, Interim Remedial Measures— Procedures (TAGM-4048), at 2-3 (Dec. 9, 1992) (providing guidance for an IRM “consist[ing] of a ‘removal’ as identified in the NCP at 40 C.F.R. § 300.415(d)” and an IRM “representing] a significant portion of the remedy,” which “must be part of the Proposed Remedial Action Plan (PRAP)/ROD process at the completion of the Remedial Investigation/Feasibility Study”).18 Clearly, IRMs can either be removal or remedial actions. Accordingly, because the 1997 cleanup was part of a larger remedial action that began in 1993, NYSEG’s claims are time-barred.
c. Owego
The district court found that the six-year statute of limitations barred NYSEG from recovering costs for the 2003 cleanup of a portion of the Owego site because that cleanup was part of a remedial cleanup begun in 1994. NYSEG, 808 F.Supp.2d at 511. NYSEG argues that the 2003 cleanup was separate from the 1994 cleanup, triggering a separate statute of limitations.
Between September 1994 and July 1995, NYSEG remediated the portion of the site known as “Operational Unit 1” (“OU-1”) by removing and thermally destroying 13,-000 tons of coal tar-contaminated soil. Id. NYSEG performed this work pursuant to the ROD issued by DEC in 1994.
In 1996, NYSEG discovered more coal tar contamination in the Susquehanna River on what became “Operational Unit 2” (“OU-2”). In 1998, NYSEG identified a pipe leading to the riverbed as the source of contamination. In 2002, DEC ordered NYSEG to remove the pipe as well as the contaminated river sediments. In 2003, NYSEG completed remediating the river and pipe by removing 1,200 tons of sediment and 30 feet of pipeline.
NYSEG now seeks to recover for the cost of groundwater monitoring and pipe and sediment removal performed at OU-2 in 2003, arguing that it was a separate remedial action. FirstEnergy argues that this cleanup is a continuation of the remedial work begun in 1994 on OU-1, and thus time-barred.
The district court found that the work at OU-2 was part of the earlier remedial action at OU-1. We agree. First, the work at OU-2 was remedial — it was designed to permanently eliminate the coal tar contamination from the Susquehanna River by removing the pipe leading to the river and the contaminated sediment.
Second, the district court correctly held that there can only be one remedial action at any given site. Id. Virtually every court that has considered this issue has agreed. See generally Colorado v. Sunoco, 337 *236F.3d 1233, 1241 (10th Cir.2003) (“In our view, this language [of the statute] indicates there will be but one ‘removal action’ per site or facility, as well as a single ‘remedial action’ per site or facility.”); Kelley v. E.I. DuPont de Nemours & Co., 17 F.3d 836, 843 (6th Cir.1994) (holding that all removal activity should be considered part of one removal action because “[i]t is simply inconsistent with the[] ‘essential purposes’ [of CERCLA] to require suit on each arguably independent removal activity”) (internal citations omitted); Yankee Gas Servs. Co., 616 F.Supp.2d at 270 (“[C]ourts have generally held that there can be only one removal and one remedial action per facility, regardless of the number of phases in which the clean-up occurs.”); California ex rel. Cal. Dep’t of Toxic Substances Control v. Hyampom Lumber Co., 903 F.Supp. 1389, 1394 (E.D.Ca.1995) (“There is no authority for the view that each ‘remedial’ activity undertaken at a site triggers a new cause of action for the cost recovery of that activity.”). But see United States v. Manzo, 182 F.Supp.2d 385, 402 (D.N.J.2000) (finding different statutes of limitation for different operable units); United States v. Ambroid Co., Inc., 34 F.Supp.2d 86, 88 (D.Mass.1999) (finding more than one removal action in the “unusual circumstance” where eighteen months after EPA closed site leaving behind drums of hazardous waste which had no legally-approved method of disposal, site was broken into and drums damaged, requiring second removal action).
We agree with the majority view and hold that there can only be one remedial action at a site. The very nature of a remedial action is to permanently remediate hazardous waste. Next Millenium, 732 F.3d at 125. A remedial action is supposed to be a final, once-and-for-all cleanup of a site, and once a PRP completes an approved remediation plan, it would not be logical — or fair — to subject that entity to additional CERCLA lawsuits seeking yet additional permanent relief, although we recognize that what seems final at a given point in time might come to appear inadequate at a later date as scientific knowledge progresses.19
This limitation on remedial actions does not, however, foreclose cost recovery actions for supplemental cleanup costs incurred after the limitations period has run on the initial cleanup, if they are filed within three years of the completion of the remedial work. Section 9613(g)(2) distinguishes between an “initial action” and a “subsequent action or actions” to recover additional response costs as part of the same remedial action. 42 U.S.C. § 9613(g)(2). Section 9613(g)(2) provides that subsequent actions for “further response costs ... may be maintained at any time during the response action, but must be commenced no later than 3 years after the date of completion of all response action.” Id. As long as the party seeking to recover costs filed an initial action for cost recovery within the time period specified in § 9613(g)(2)(A) and (B), the party can recover the costs of later cleanups if the action to recover such costs is filed no later than three years after the date of completion of the new cleanup. See id. Here, the later cleanup could not qualify as a “further response cost” because the three-year period had expired. Id.
Therefore, the district court correctly held that NYSEG’s claims for cost recovery for the remedial work at OU-2 is time-barred.
*237 5. Gas Production
The district court used total gas production at each site to allocate liability because both parties agree that coal tar was generally created in proportion to the amount of gas produced. NYSEG, 808 F.Supp.2d at 484. The parties disagree, however, as to how to determine the amount of gas produced by a particular plant at a particular time. While the parties agree on the amount of coal tar produced between 1887 and 1932, they do not agree about the amount of coal tar produced before 1887 or after 1932.20
Specifically, NYSEG argues that the district court was incorrect in adopting First-Energy’s estimates for the pre-1887 and post-1932 period. Regarding the pre-1887 period, FirstEnergy’s expert took the production numbers reported in the first year of the Brown’s Directory in 1887 for a particular plant and assumed that the plant had produced one-half of that amount of gas each year, from its first year of operation until 1887.
For the period from 1932 to 1940, First Energy’s expert used Brown’s Directory records which contained system-wide (not plant-by-plant) reports for 1932 to 1940. He then made production estimates for each plant based on those records and extrapolated to estimate production for years when exact gas volume data was not available.
NYSEG argues that for the period before 1887, the district court should have found no gas production at all — even though there was gas production — because the numbers are not reliable. For the 1932 to 1940 period, NYSEG’s expert relied on yearly production figures found in data collected by the PSC for most of the operating plants. If a PSC record was not available for that year, NYSEG’s expert treated that year as if there were no gas production.
In light of the severely limited information available, the district court did not err in using FirstEnergy’s estimates. It was reasonable for the court to use its best estimates for periods in which there was indisputably some gas production, rather than to treat those years as if there were no production at all.
6. NYSEG’s $20 Million Insurance Settlement
NYSEG received $20 million in an insurance settlement, which covered thirty-eight sites, including the sixteen sites in this suit. Id. at 484. The settlement also extinguished third-party property damage claims against the settling insurer. Id.
Although there was no evidence regarding the allocation of the settlement amount among the thirty-eight sites, the district court chose to apply a portion of the settlement to reduce FirstEnergy’s cleanup costs. Id. at 527. In doing so, the district court calculated a -pro rata share of the insurance settlement by using 42.1 percent of $20 million, which represented the proportion between the sixteen sites in this case and the thirty-eight involved in the insurance settlement. Id. This came to a total of $8,421,052, the pro rata share of the settlement attributable to the sixteen sites. That amount is 8.9 percent of the total amount of money sought for cleanup, $94,277,153, which percentage the court then applied evenly over the sixteen sites to reduce NYSEG’s recovery. Id.
NYSEG argues that none of the money should have been applied to reduce First-Energy’s payments because (1) First-*238Energy did not carry its burden to prove the affirmative defense that the insurance proceeds should be applied to reduce its liability and (2) there is no risk of double recovery for NYSEG.21
The district court did not err in reducing NYSEG’s recovery to account for the insurance proceeds. To begin, we note that the district court was correct that the collateral source rule does not apply in CERCLA cases. The collateral source rule generally provides that “damages recoverable for a wrong are not diminished by the fact that the party injured has been wholly or partially indemnified for his loss by insurance effected by him and to the procurement of which the wrongdoer did not contribute.” Ocean Ships, Inc. v. Stiles, 315 F.3d 111, 116 (2d Cir.2002) (internal quotations marks omitted). “[T]he policy underlying the collateral source rule—to provide the innocent party with the benefit of any windfall—is simply not advanced in [CERCLA] cases.” Friedland v. TIC-The Indus. Co., 566 F.3d 1203, 1207 (10th Cir.2009); see also Basic Mgmt. Inc. v. United States, 569 F.Supp.2d 1106, 1123-24 (D.Nev.2008) (refusing to apply collateral source rule to insurance proceeds because “CERCLA contribution actions are not injury actions in which the injured party is seeking compensation for damages to be made whole again”). Rather, in resolving contribution claims under CERCLA, a “court may allocate response costs among liable parties using such equitable factors as the court determines are appropriate.” 42 U.S.C. § 9613(f)(1); see Goodrich Corp. v. Town of Middlebury, 311 F.3d 154, 168 (2d Cir.2002) (“[T]he court must allocate response costs among liable parties in an equitable manner.”).
Here, in reducing NYSEG’s recovery in light of the insurance payment, the district court reasonably exercised its equitable powers to allocate response costs between the parties. First, insofar as NY-SEG contends that FirstEnergy failed to carry its burden of proving that consideration of the insurance settlement was warranted, the argument fails. FirstEnergy presented evidence of the $20 million insurance payment. NYSEG, the party in position to explain how the proceeds were allocated among the thirty-eight sites, failed to present any such evidence. Given the sparseness of the record as to the allocation of the proceeds, therefore, the district court’s pro rata allocation based on the number of sites should be applied was not arbitrary or haphazard, but reasonable.
Second, while it is undoubtably correct that there was no risk of NYSEG receiving a double recovery, that fact did not preclude the district court from equitably reducing NYSEG’s recovery to account for the insurance proceeds. See, e.g., Basic Mgmt. Inc., 569 F.Supp.2d at 1125 (“Equity and common sense ... dictate that [p]laintiffs cannot recover the remediation costs paid for by their insurance policies.”). Instead, CERCLA’s “expansive language ... affords a district court broad discretion to balance the equities in the interests of justice.” Bedford Affiliates, 156 F.3d at 429. Accordingly, the district court did not err in applying NYSEG’s insurance settlement to reduce FirstEnergy’s liability.
7. Should NYSEG’s Recovery be Reduced Either Because of Economic Beneñt or Alleged Delay in Cleaning Up?
FirstEnergy argues that the district court should have reduced NYSEG’s re*239covery to reflect the fact that NYSEG would benefit from the increased value of the cleaned up sites. It also argues that NYSEG’s recovery should be reduced because NYSEG was slow to remediate and “engaged in unnecessarily piecemeal remediation.” FirstEnergy Br. at 87. These arguments are not persuasive.
Regarding the economic benefit of the cleanup, the district court recognized the substantial amount of money NYSEG was spending on cleanup — $94 million to date with another $114 million expected in future costs. Considering that NYSEG does not expect to collect a comparable amount from other PRPs, NYSEG is hardly gaining a windfall. FirstEnergy also fails to offer evidence about any increase in the value of the land, much less an increase over the $200 million NYSEG will expend to clean up the properties.
FirstEnergy’s argument that NYSEG caused unreasonable delay in the cleanup fails as well. FirstEnergy points to two specific examples of alleged delay: (1) in 1983, a NYSEG employee proposed a program under which all MGP investigation would be completed by 1988, with remediation completed within two years (this plan was not adopted) and (2) in 1989, NYSEG employees proposed a “comprehensive approach” to site cleanup, but NYSEG did not move forward with broad remediation efforts until 1994. FirstEnergy Reply Br. at 35. FirstEnergy offers no explanation as to why NYSEG was unreasonable in failing to adopt the 1983 plan, nor has it shown that the plan was even technically or financially feasible. FirstEnergy also fails to explain how NYSEG was unreasonable in adopting the 1994 plan instead of the 1989 proposal. In short, the record simply does not support FirstEnergy’s argument that NYSEG caused unreasonable delay in the cleanup by declining to follow a plan suggested in 1983 or delaying its cleanup plan until it adopted a joint plan with DEC in 1994.
8. I.D. Booth
I.D. Booth bought the Cortland-Homer MGP from Mack Trucks, Inc. in 1971, which bought it from NYSEG. NYSEG, 808 F.Supp.2d at 453. As a result, it is liable under CERCLA as a current owner. It now raises two issues on appeal. First, I.D. Booth argues that the district court erred in finding it liable for cleanup costs at the Cortland-Homer site. Second, I.D. Booth argues alternatively that the amount of liability assigned to it by the district court was too high.
The Cortland-Homer site is divided into two parcels: OU-1, which is comprised of the former MGP area and the structures below its surface, and OU-2, which consists of the adjacent land. I.D. Booth’s office building was located in OU-1 and sat on top of two buried gasholders.
Sometime in the mid-1990s, after coal tar contamination was discovered on the Cortland-Homer site, NYSEG offered to buy the southern portion of the Booth building from I.D. Booth. Although the building was appraised at $350,000, I.D. Booth demanded $2 million, which included the cost of business disruption. Negotiations began in mid-November 2005. On May 8, 2008, NYSEG paid $1.8 million for the southern portion of the building and gave I.D. Booth the right to buy the property back after remediation for $1 in the event NYSEG were to decide to sell it. I.D. Booth retained ownership of the northern portion of the building.
I.D. Booth claims to be entitled to a “third-party” defense under section 107(b)(3), which requires it to establish that:
[T]he offending “release ... of a hazardous substance and the damages result*240ing therefrom were caused solely by ... an act or omission of a third party,” provided that (1) the third party is not “one whose act or omission occurs in connection with a contractual relationship, existing directly or indirectly, with the defendant,” (2) the defendant “took precautions against foreseeable acts or omissions of any such third party and the consequences that could foreseeably result from such acts or omissions,” and (3) the defendant “exercised due care with respect to the hazardous substance concerned, taking into consideration the characteristics of such hazardous substance, in light of all relevant facts and circumstances.”
New York v. Lashins Arcade Co., 91 F.3d 353, 359 (2d Cir.1996) (quoting 42 U.S.C. § 9607(b)(3)). I.D. Booth bears the burden of showing by preponderance of the evidence that all four of these conditions are met. NYSEG, 808 F.Supp.2d at 514.
Here, the district court correctly held that I.D. Booth failed to meet its burden with respect to the third requirement because it did not exercise due care with respect to the cleanup effort. Id. at 519; see Lashins Arcade Co., 91 F.3d at 361 (explaining that “due care would include those steps necessary to protect the public from a health or environmental threat”) (internal citations and quotations omitted).
The district court found that I.D. Booth’s tactics created extensive delays in negotiations. Despite knowing that NY-SEG could not begin remediation until it acquired portions of the Cortland-Homer site — which included the removal of two former gasholders from below I.D. Booth’s building — I.D. Booth engaged in “protracted negotiations.” NYSEG, 808 F.Supp.2d at 518. As a result, I.D. Booth was largely responsible for a two-year delay in the cleanup due to its untimely responses to NYSEG’s proposals and its “aggressive price demand.” Id. As FirstEnergy’s expert testified, this delay most likely allowed more coal tar to migrate. Id. at 519. Because I.D. Booth’s delays prevented NYSEG from obtaining control of the I.D. Booth building located on OU-1, NY-SEG was forced to consider implementing the remedy at OU-2 first, even though this was not the optimal order in which to proceed with cleanup. Id.
To further exacerbate the problem, I.D. Booth’s cooperation in cleanup efforts was “somewhat lackluster,” as it failed, for example, to provide NYSEG with “feedback.” Id.; see Lashins Arcade Co., 91 F.3d at 361 (landowner failed to exercise “due care” where it remained passive “simply because public environmental authorities are addressing a hazardous waste situation”). Because this finding was not clearly erroneous, we affirm.22
I.D. Booth also complains that the portion of the cleanup costs assigned to it was too high. The district court assigned I.D. Booth 15 percent of FirstEnergy’s costs at the Cortland-Homer site, or 6.7 percent of NYSEG’s total Cortland-Homer costs. NYSEG, 808 F.Supp.2d at 535, 542. Accordingly, the district court found that I.D. Booth was liable for $160,089 in past costs, and for 15 percent of all future “necessary and NCP compliant costs incurred by NY-SEG.” Id. I.D. Booth argues that even if it caused a delay, that delay should be *241measured from date of the issuance of the ROD approving the remediation method (March 2007) to the date of sale (April 2008), or 14 months. Using those dates, I.D. Booth would only owe 1.35 percent of total costs — 14 months out of 75 total years of production (1858 to 1933). In the alternative, I.D. Booth argues that the amount it owes should only be .67 percent, which is one year out of the roughly one-hundred and fifty years the site has been in existence (1858-2014).
We find that the district court did not abuse its discretion in allocating the response costs. The district court reasonably took into account the fact that I.D. Booth would benefit from the increased property value after remediation and that its negotiating tactics led to significant delays in remediating the property. Id. at 533. Accordingly, we affirm.
CONCLUSION
In sum, we hold:
(1) NYSEG’s CERCLA claims against FirstEnergy are not barred by the covenant not to sue;
(2) AGE CO is not directly liable under CERCLA as an operator;
(3) FirstEnergy is liable to NYSEG on a veil piercing theory based on AGECO’s control over NYSEG from 1922 to January 10, 1940, but not for contamination created by other AGE CO subsidiaries before those subsidiaries merged into NYSEG;
(4) NYSEG’s claims as to the (a) Platts-burgh site are timely, (b) Norwich site are untimely, and (c) Owego site are untimely;
(5) The district court did not err in calculating total gas production at the sites;
(6) The district court did not abuse its discretion in reducing NYSEG’s recovery from FirstEnergy by a portion of NY-SEG’s $20 million insurance settlement;
(7) The district court did not abuse its discretion in declining to reduce NYSEG’s recovery to reflect the increased value of the remediated properties or NYSEG’s alleged delay in the remedial efforts; and
(8) I.D. Booth is liable for a portion of cleanup costs and the district court did not abuse its discretion in apportioning liability in this respect.
For the foregoing reasons, the judgment of the district court is AFFIRMED in part and VACATED in part, and the case is REMANDED for further proceedings.