concurring in part and dissenting in part:
I concur in holding that EPA impermis-sibly considered costs in the rule on review. See Op. Part II. I write separately in Part I on the nature of the error, because I cannot join the court’s reasoning, and in Part II on the appropriate remedy, because the court has ignored binding precedent. Thereafter, in light of the required remand, I address in Part III Honeywell’s contention that EPA’s past approval of other substitute chemicals as safe alternatives preclude it from making any exceptions authorizing use of HCFC-*137622 or HCFC-142b for some foam end users.
I.
The challenged regulation permits new use of HCFC-22 and HCFC-142b only when “necessary to meet performance or safety requirements,” and “technical constraints [ ] preclude the use of other available substitutes.” 40 C.F.R. Pt. 82 Subpt. G, app. K (2004). Whether EPA imper-missibly took costs into consideration depends on what it means for “technical constraints” to “preclude” the use of other available substitutes. While there is language in the rule indicating that EPA considered costs, the more natural reading of the regulation, in light of the language EPA used, is, as EPA counsel maintains, that the regulation means what it says it means: an end-user may only use HCFC-22 or HCFC-142b if it is not actually possible to use anything else. The fact that it might be more expensive to use an approved alternative would not suffice — use of the hydrochlorofluorocarbon must be “necessary.”
This reading of the regulation is certainly natural, as it gives the term “preclude” its most obvious meaning of “to make impossible.” AmeRican Heritage Dictionary of the English Language 212 (4th ed.2000). Indeed, absent EPA’s reference in the final rule to costs, the court would likely read the regulation as EPA’s counsel suggests, giving the word “preclude” its most natural, dictionary meaning. Cf. Citizens Against Burlington, Inc. v. Busey, 938 F.2d 190, 202-02 (D.C.Cir.1991). If technical constraints must “make impossible” the use of an approved substitute before use of a hydrochlorofluorocarbon is permitted, then the court’s concern about EPA permitting HCFC-22 or HCFC-142b use on the basis of mere convenience or cost, see Op. at 1372-73, would be inapposite: modifications might be required to shift to an approved substitute that “make difficult” or “make costly” or “make inconvenient” the use of an approved substitute, but they would not “preclude” it. Notably, the record before the court indicates that the transition from HCFC-141b to HCFC-22 or HCFC-142b is more likely to involve equipment modifications than the transition from HCFC-141b to an approved alternative. HFC-245fa, Honeywell’s product, is a liquid agent like HCFC-141b, and was developed specifically as a “drop-in” substitute for use in the same equipment. HCFC-22 and HCFC-142b, on the other hand, are both gaseous agents, which, EPA acknowledged, would often require modifications to equipment currently using liquid agents such as HCFC-141b. See 67 Fed.Reg. 47,703, 47,713-14 (July 22, 2002). Some transition costs, then, would cut against HCFC-22 and HCFC-142b rather than in their favor. Thus, EPA counsel’s suggestion, that the HCFCs’ use under the new regulation is intended for cases where foam made with an approved substitute would not be functional, is not without plausibility.
Under a natural reading of the language used by EPA, HCFC-22 and HCFC-142b use would be reserved, not for situations where it would be “possible to fit a round peg in a square hole” with sufficient expense, Op. at 1372, but for situations where the physical limitations of approved foamblowing (or foam-pouring) agents are such that equipment or design modifications will not be sufficient to meet “performance or safety requirements.” With this view of EPA’s meaning, the fact that EPA also referred to costs in explaining the regulation would be harmless error. Cf. PDK Laboratories Inc. v. DEA, 362 F.3d 786, 799 (D.C.Cir.2004); 5 U.S.C. § 706. The bulk of EPA’s explanation of the regulation in the final rule is consistent *1377with this natural reading of the regulation. EPA relied chiefly on its concern that foam made with existing alternatives might, in some instances, not meet certain dimensional, flammability control, and insulation requirements. See 67 Fed.Reg. at 47,713-14. EPA did not conclude that approved alternatives would produce less desirable foam; rather, it expressed concern that it did not yet have enough information to confirm whether the foam made with existing approved substitutes to HCFC-141b would always work functionally as well as foam made with HCFC-22 or HCFC-142b.
The court appears to hold that EPA impermissibly considered costs at two stages of its decision: first, by comparing the functionality of foam made with approved alternatives with foam made with HCFC-22 and HCFC-142b, and second, by considering the costs manufacturers would incur in switching to approved alternatives. See Op. at 1372-73. While EPA relied on the first consideration, it was not error to do so. The court defines too broadly what it means for EPA to imper-missibly consider costs. Expressing concern over whether a substitute product actually works (or works as well) as the substance it is replacing is, of course, a decision that may carry economic consequences, as where a less functional foam product will be less commercially desirable. If a foam is denser, picnic coolers will have to be heavier to keep the same amount of food cold; if a foam is less insular, it will require thicker walls in refrigerators or houses that use it as insulation. But that does not convert every decision EPA makes about whether a substitute works into a decision about costs: were that so, every approval or disapproval decision EPA makes under the SNAP program would be suspect, an extreme position Honeywell does not urge.
The court’s second theory, however — that EPA based its decision in part on the costs manufacturers would incur in switching to approved alternatives — is indicated by the record. See Op. at 1372-73. A statement by EPA in the rule, about protecting small businesses, calls into question the natural reading of the regulation urged by EPA counsel:
In other cases, where HCFC-141b is used in niche applications, EPA believes foam manufacturers may experience difficulties and delays in transitioning from HCFC-141b to non-ozone-depleting alternatives. Given the constraints associated with cost and timing of transitioning to alternatives for small businesses, and the need to facilitate a smooth and equitable transition from HCFC-141b, EPA believes ... it is appropriate to approve use of HCFC-22 and/or HCFC-142b as substitutes for HCFC-141b ... provided that the users ... ascertain and document that other acceptable alternatives are not technically feasible.
67 Fed.Reg. at 47,714 (emphasis added). It is difficult to understand this passage unless EPA believes there is some subset of end-users for whom it would be possible yet very costly to switch to non-ozone-depleting alternatives, and that the rule grants this subset some form of relief. This, in turn, suggests that EPA construes the term “preclude” to mean something less than “make impossible,” such as to “make difficult” or “make cumbersome.”1 *1378If so, EPA’s approach would appear inconsistent with its current SNAP regulations.
Section 612(c) directs the Administrator of EPA to make the use of environmentally harmful substitutes unlawful if, inter alia, alternatives presenting reduced harm are “currently or potentially available”; the regulations provide that a substitute will be deemed “unacceptable” if “other alternatives exist that reduce overall risk.” 40 C.F.R. § 82.180(b)(4) (emphasis added). The court, in dicta, opines that a substitute may not be “currently or potentially available” within the meaning of CAA § 612(c) if using it would be too impractical, see Op. at 1373, and hence, presumably, does not “exist” for purposes of § 82.180(b)(4). But the SNAP regulations nowhere attempt to make this link. The definition of “potentially available” turns only on whether EPA “reasonably believes” a substitute “to be technically feasible,” id. § 82.172, and substitutes must be designated “unacceptable” if cleaner alternatives “exist.” Id. at § 82.180(b)(4). While the SNAP regulations make the “cost and availability of the substitute” an element of acceptability, id. § 82.180(a)(7)(vii), that concern is limited to whether EPA “has ... reason to prohibit its use,” id. § 82.180(b)(1), not to whether cleaner alternatives for the substance are already “currently or potentially available.” Under the SNAP regulations the fact that it might be difficult or time-consuming for some small businesses or others to use “other alternatives” is irrelevant, so long as those alternatives “exist.” Consideration of transition costs is thus precluded by the SNAP regulations as currently written, irrespective of whether it might be permitted under CAA § 612(c), and it is arbitrary and capricious for EPA to fail to comply with its own regulations. See Dithiocarbamate Task Force v. EPA, 98 F.3d 1394, 1398-1402 (D.C.Cir.1996).
For present purposes, as the court states, see Op. at [1365-66], there is no need to decide whether EPA could, consistent with the text and purpose of CAA § 612, adopt SNAP regulations that incorporate, into a determination of whether a cleaner substitute is “currently or potentially available,” consideration of the cost and practicality of transition. While the court speculates that such regulations might pass muster, it properly acknowledges that this question is not before the court. See Op. at 1373. The SNAP regulations currently include no such provision, and EPA did not attempt, either in the rule or in its brief, to justify relying on costs in the rule on the basis that § 612(c) permits it to do so. It remains to be seen whether EPA will decide to make consideration of transition costs an element of the SNAP process.
Whether CAA § 612(c) would permit substantive consideration of transition costs is not apparent on the face of the statute and presents a serious question of statutory interpretation. Heretofore, when Congress has wanted the Administrator to consider costs under the CAA it has expressly called for consideration of costs or practicality. See, e.g., 42 U.S.C. §§ 7411(a)(1), 7412(d)(2), 7479(2)(C)(3). Title Vi’s policy of phasing out ozone-*1379depleting substances involves express consideration of the practicality of transition, but at a different point in the process: in deciding whether an accelerated phase-out schedule is warranted, the Administrator is to consider whether “such more stringent schedule is practicable, taking into account technological achievability, safety, and other relevant factors.” CAA § 606(a)(2), 42 U.S.C. § 7671e (emphasis added). The decision that an accelerated phaseout of HCFC-141b is practicable has already been made. See 58 Fed Reg. 65,-018, 65,028 (Dec. 10, 1993). The court has repeatedly held in cases involving other sections of the CAA that cost plays no role in the promulgation of emissions standards. See, e.g., American Lung Assn. v. EPA 134 F.3d 388, 389 (D.C.Cir.1998); NRDC v. Adm’r, EPA 902 F.2d 962, 973 (D.C.Cir.1990), vacated in part on other grounds, NRDC v. EPA 921 F.2d 326 (D.C.Cir.1991); American Petroleum Inst. v. Costle, 665 F.2d 1176, 1185 (D.C.Cir.1981); Lead Indus. Assn., Inc. v. EPA 647 F.2d 1130, 1148 (D.C.Cir.1980). The Supreme Court’s decision in Whitman v. American Trucking Ass’n, 531 U.S. 457, 467-70, 121 S.Ct. 903, 909-10, 149 L.Ed.2d 1 (2001), which rejected a reading of the term “public health” in the CAA that incorporated cost considerations, further cautions against reading economic considerations into the CAA where they do not appear on the face of the statute. Thus, whether CAA § 612(c) might permit consideration of practicality in extreme cases, such, as the court hypothesizes, where it would be so difficult to “fit a round peg in a square hole,” Op. at 1372, that a non-ozone-depleting alternative could no longer be said to be “available,” is a question that is not yet before the court. EPA has not attempted to locate its approach in the statutory text, and it behooves the court, in light of the deference that may be due, to afford EPA the opportunity to decide whether transition costs are to be considered in evaluating a clean alternative’s availability.
II.
While the court properly identifies error in the regulation on review, it goes astray on the appropriate remedy. Under the court’s precedent, the proper remedy in this case is remand, not vacatur. The court explained in Allied-Signal v. U.S. Nuclear Regulatory Comm’n, 988 F.2d 146, 150-51 (D.C.Cir.1993), that remand is preferable to vacatur when an agency might be able to support a rule through further explanation and the “consequences of vacating may be quite disruptive.” Id. at 151. See also Sinclair v. FCC, 284 F.3d 148, 162 (D.C.Cir.2002); Fox Television Stations, Inc. v. FCC, 280 F.3d 1027, 1047-49 & 1052-53 (D.C.Cir.2002). The two Allied-Signal factors are both implicated here.
First, the identified defect in the rule that the court has identified may be curable. As discussed, the final regulation permits new use of HCFC-22 and HCFC-142b in certain end-uses when “necessary to meet performance or safety requirements,” and “technical constraints [] preclude the use of other available substitutes.” 40 C.F.R. Pt. 82 Subpt. G, app. K. EPA’s counsel suggested a permissible reading of the regulation based on the natural meaning of its terms: that the chemicals may only be used when it is impossible, from a technical standpoint, to use anything else. At the same time, EPA’s explanation that it allowed new uses because of “the constraints associated with cost and timing of transitioning to alternatives for small businesses,” 67 Fed.Reg. at 47,714, suggests a policy of protecting small businesses from the costs of a possible, yet difficult, transition, and thereby permitting the use of HCFC-22 and *1380HCFC-142b even in some cases where available substitutes already “exist.” Cf. 40 C.F.R. § 82.180(b)(4). But if EPA intended no more than the natural reading, it can say so on remand, in which event any impermissible reference to costs in the final rule was irrelevant to the regulation actually promulgated, and thus harmless error. Cf. PDK Laboratories, 362 F.3d at 799; 5 U.S.C. § 706. Of course, it is possible that EPA may instead decide to promulgate new SNAP regulations permitting consideration of transition costs in determining the availability of alternatives, or it may promulgate a new rule in which cost is not a guiding consideration, based on its existing SNAP regulations. Yet because EPA might wish to retain the rule on review, and might be able to easily cure any defect, vacating the rule risks unnecessary disruption to the regulatory scheme. Cf. Allied-Signal, 988 F.2d at 150-51; Davis Cty. Solid Waste Mgmt. v. EPA 108 F.3d 1454 (D.C.Cir.1997) (per curiam).
Second, were the court to vacate the final rule with respect to the end-uses in which EPA has permitted limited new HCFC-22 and HCFC-142b use, the result would likely be that all uses of HCFC-22 and HCFC-142b within those end-uses would become permissible, as more than 90 days have elapsed since ATOFINA’s petition and the chemicals were not listed as unacceptable in those end-uses prior to the proposed rule. See 40 C.F.R. § 82.174(a). This would likely lead to more widespread interim use of these environmentally harmful chemicals than the much more limited use EPA authorized in the rule on review. In the past, where vacatur of a rule risks interim harm to the environment, the court has allowed successfully challenged EPA rules to remain in effect pending remand. See Davis County Solid Waste Mgmt., 108 F.3d at 1459-60. Given the possibility of confusion and environmental harm, the same course of action is warranted here.
Rather than engage the prudential inquiry our case law requires to determine whether remand or vacatur is the proper course of action, the court sua sponte holds that there is no statutory authority for the court to remand without vacating. See Op. at 1373-74. It brushes aside our exhaustive caselaw on this question, much of it quite recent, see, e.g., Milk Train, Inc. v. Veneman, 310 F.3d 747, 755-56 (D.C.Cir.2002); Fox Television Stations, 280 F.3d at 1047-49 & 1052-53; County of Los Angeles v. Shalala, 192 F.3d 1005, 1023 (D.C.Cir.1999); Radio-Television News Dirs. Ass’n v. FCC, 184 F.3d 872, 887-89 (D.C.Cir.1999); Allied-Signal, 988 F.2d at 150-51, by claiming that these cases do not actually constitute precedent and are, in any event, distinguishable. See Op. at 1374. Neither argument is persuasive.
The court rewrites the law of precedent, allowing a later court to avoid binding decisions it prefers not to follow if the earlier decisions did not fully state their reasoning. See Op. at 1374. But it is long established that irrespective of whether a later court may conclude that its view is the better view, it is bound, absent en banc review, by the court’s prior decisions. See LaShawn A. v. Barry, 87 F.3d 1389, 1395 (D.C.Cir.1996) (en banc); 28 U.S.C. § 46(c) (2004). It is true that questions that “lurk in the record; neither brought to the attention of the court nor ruled upon” do not constitute precedent. Webster v. Fall, 266 U.S. 507, 511, 45 S.Ct. 148, 69 L.Ed. 411 (1925). But the question the court revives today- — the court’s statutory authority to remand agency action without vacating — has been “brought to the attention of the court” on at least two occasions by separate opinions, and on both occasions, the court remanded the agency action in ques*1381tion without vacating, over separate opinions raising substantially the same arguments as those that make up the holding of the court today. See Milk Train, 310 F.3d at 757-58 (Sentelle, J., dissenting); Checkosky v. SEC, 23 F.3d 452, 491 (D.C.Cir.1994) (per curiam) (separate opinion of Randolph, J.).
Nor can the court reasonably distinguish our precedent by pointing out that most of it arose under the Administrative Procedure Act (“APA”) rather than the CAA, and that the CAA only grants the court the power to “reverse” administrative actions. 42 U.S.C. § 7607(d)(9)(C) (2004). The court does not explain how the relevant language in the CAA is any different from § 706 of the APA, which confers the power to “set aside” agency action, 5 U.S.C. § 706(2), where the court’s opinion itself argues that “set aside” is a synonym for “reverse.” Op. at 1374 (quoting Web-steR’s New Collegiate Diotionaky 991 (1973)). And the court’s attempt to distinguish Davis, where the court, applying the Allied-Signal test, remanded a rule under the CAA without vacating, 108 F.3d at 1459-60, by stating that the parties in Davis did not raise the question of the court’s authority to remand, Op. at 1374, ignores that the parties did not raise the issue in the instant case either. Judge Randolph’s approach, see Concurring Op. at 1374-75, may be of interest to the en banc court, but, absent en banc review, our precedent requires that the rule be remanded to EPA for further explanation.
III.
Finally, in light of the required remand, I address Honeywell’s contention that it is entitled to greater relief than the court grants. As discussed, EPA’s decision to approve some limited use of HCFC-22 and HCFC-142b was based in large part on its concern that it lacked sufficient information on whether existing “safe alternatives” for HCFC-141b could always function adequately as substitutes. Honeywell contends that this inquiry was foreclosed by EPA’s past approval of safe alternatives to HCFC-141b and past refusal to approve ozone-depleting substances as safe alternatives, and that EPA was thus precluded from approving any ozone-depleting substances as substitutes for HCFC-141b, in any end-uses. However, irrespective of whether CAA § 612(c) would permit EPA to approve some limited use of HCFC-22 and HCFC-142b on remand, neither of these other objections present separate obstacles to such approval.
Under CAA § 612(c), the Administrator is required to “publish a list of ... safe alternatives identified under this subsection,” and to promulgate regulations making it unlawful to use any substitute “which the Administrator determines may present adverse effects to human health or the environment, where the Administrator has identified an alternative to such replacement that — (1) reduces the overall risk to human health and the environment; and (2) is currently or potentially available.” Id. Honeywell maintains that EPA violated this provision because it found that suitable alternatives to HCFC-141b were not “potentially available” even though it had already designated several such substitutes as “safe alternatives.” HCFC-22 and HCFC-142b are chemicals that EPA has determined to “present adverse effects to ... the environment”; the importation and manufacture of both was banned, effective 2010, in the same rule in which EPA imposed the 2003 ban on the importation and manufacture of HCFC-141b. See 58 Fed.Reg. at 65,028. EPA therefore cannot permissibly approve their use as substitutes if the Administrator has “identified an alternative” that “(1) reduces the overall risk to human health and the *1382environment; and (2) is currently or potentially available.” CAA § 612(c).
In Honeywell’s view, EPA’s approval in 1999 of HFC-245fa in “all foam end-uses,” 64 Fed.Reg. 68,039, 68,041 (Dec. 6, 1999), as well EPA’s approval of other non-ozone-depleting substitutes, see 67 Fed.Reg. at 47,711, Table B, encompassed a determination that it is “technically feasible” to use HFC-245fa in all foam end-uses, and that it is therefore “currently or potentially available.” This contention conflates two terms in § 612(c): an “alternative ... that is potentially available” and a “safe alternative[] identified under this subsection for specific uses.” Being a “safe alternative” and being “currently or potentially available,” under EPA’s regulations, are not the same thing. The SNAP regulations governing approval of substitutes call only for applicants to identify “applications within each sector end-use in which the substitutes are likely to be used,” 40 C.F.R. § 82.178(a)(3), and provide that EPA will approve a substitute where it has “found no reason to prohibit its use,” id. § 82.180(b)(1), with the decision based on “[a]tmospherie effects,” “[g]eneral population risks,” “[ejcosystem risks,” “[o]ecupa-tional risks,” “[consumer risks,” “[f]lamm-ability,” and “[c]ost and availability.” Id. § 82.180(a)(7). On the other hand, “potentially available” is defined as “any alternative for which adequate health, safety, and environmental data, as required for the SNAP notification process, exist to make a determination of acceptability, and which the Agency reasonably believes to be technically feasible, even if not all testing has yet been completed and the alternative is not yet produced or sold.” Id. § 82.172. In other words, being “potentially available” adds the additional element of EPA’s reasonable belief in technical feasibility that is not required of substances designated as safe alternatives. A substance can be a “safe alternative” for a particular end-use and yet be “potentially available” for only some, not all, applications within that end-use. EPA was therefore free to decide in this rulemaking proceeding whether specific applications remained in which permissible constraints still prevented the use of existing, non-ozone depleting substitutes such as HFC-245fa, thus rendering them not “potentially available” for those uses. In light of the remand, it is unnecessary to decide whether EPA had adequate record support for its conclusion that such constraints still existed, but EPA’s past approval decisions did not prevent it from evaluating the feasibility of existing alternatives.
Honeywell’s related contention that EPA impermissibly deviated from its stated policy of never approving one ozone-depleting substance as a substitute for another fails for essentially the same reason. In prior rules where EPA had expressed this position, such as in its 1999 refusal to list NARM-22, a blend containing HCFC-22, EPA had always qualified it by noting that “[o]ther alternatives ... are already acceptable that do not contain any ozone-depleting refrigerants.” 64 Fed.Reg. 22,-982, 22,984 (Apr. 28, 1999). Similarly, in 1993, EPA stated that it “views HCFCs as important interim substitutes that will allow for the earliest possible phaseout of CFCs and other Class I substances” and “believes that the use of HCFCs should be limited to only those applications where other environmentally acceptable alternatives do not exist.” 58 Fed.Reg. at 65,026. The informal expressions of EPA policy at recent trade shows, to which Honeywell refers, similarly contain the qualification that EPA is “unlikely” to approve any HCFCs as substitutes for HCFC-141b because, among other things, there are “non-ozone depleters available.” A1 these amount to are a series of situations where EPA found non-ozone-depleting “safe al*1383ternatives” to be “available;” in this particular instance EPA determined that the existing “safe alternatives” were not “available” because some specific niche applications remained in which constraints might currently prevent the use of existing, non-ozone depleting substitutes such as HFC-245fa. Whether or not the record supports this conclusion, it was not a departure from any past policy of not approving ozone-depleters.
Honeywell can fare no better with its other contentions, although I do not reach the merits of either. The relief granted by the court eliminates the need to address Honeywell’s contention that it did not have adequate notice of EPA final decision. And Honeywell’s contention that the final rule’s delegation of technical feasibility determinations to the end-user is contrary to CAA § 612(c)’s requirement that the “Administrator” identify 'substitutes and to EPA’s regulations defining chemicals as “potential available” if “the Agency reasonably believes them to be technically feasible,” 40 C.F.R. § 82.172 (emphasis added), comes too late. Honeywell waited to make this argument in its Reply Brief, contending that the delegation undermines the narrowed use limitations by creating the risk that end-users will impermissibly consider transition costs as a basis for determining that switching to ozonefriendly alternatives is not technically feasible. See Pet. Reply Brief 2, 12. Although such delegation might present a potentially serious issue, cf. U.S. Telecom Ass’n v. FCC, 359 F.3d 554, 573-74 (D.C.Cir.2004), the court has “repeatedly held that an argument first made in a reply brief ordinarily comes too late for our consideration” because opposing counsel is denied an opportunity to respond. See Amgen v. Smith, 357 F.3d 103, 117-18 (D.C.Cir.2004) (quoting Students Against Genocide v. Dep’t of State, 257 F.3d 828, 842 (D.C.Cir.2001)). Because EPA has had no opportunity to address the delegation issue, and might promulgate a new regulation on remand that does not include the narrowed use limits, its resolution must await another day.
. To make the same point, the court also cites EPA's decision to permit existing users of HCFC-22 and HCFC-142b to continue their use of those chemicals notwithstanding the existence of non-ozone-depleting alternatives to HCFC-22 and HCFC-142b, in large part because of the costs they would incur were they to use alternatives. See Op. at 1372-73. *1378EPA quite expressly relied on transition costs in making that decision. However, EPA's decision to allow continued use of HCFC-22 and HCFC-142b is not challenged by Honeywell, which petitions for review only of EPA's decision to allow new uses of those chemicals, by users previously using HCFC-141b. Given the equipment modifications required to shift from a liquid to a gaseous foam-manufacturing agent, die fact that EPA considered transition costs as a basis for grandfathering existing HCFC-22 and HCFC-142b users does not reveal whether such costs played a role in EPA's decision to permit new uses.