On these appeals, certain pharmacies, several organizations representing pharmacies and an organization representing pharmacy benefit managers (“PBMs”) challenge settlements in two class actions. The actions were brought by purchasers of pharmaceutical drugs against publishers of drug pricing data, as well as a wholesaler not party to these appeals. At issue are both the validity of the settlements approved by the district court and the right of the would-be appellants themselves to contest it on appeal.
In the district court, class actions on behalf of drug purchasers were brought against the two publishers, First DataBank and Medi-Span, as well as McKesson — a major drug wholesaler that also owns pharmacy-related businesses. The class was comprised of third-party payors (“TPPs”) — such as health insurers like Blue Cross — who pay for all or a portion of the cost of drugs used by their beneficiaries; also included were consumers whose “co-payments” were calculated as a percentage of the price of the drugs, and consumers who were un-insured and paid for drugs out-of-pocket.
Some PBMs may have been included in the plaintiff class. TPPs often contract with PBMs to handle contracting with pharmacies, monitoring and related administrative tasks needed to have the pharmacies supply TPP beneficiaries; depending on its agreement, the PBM may act as an agent for the TPP or on its own behalf. The class in this case excluded PBMs except where the PBM was the TPP’s fiduciary or bore the insurance risk of the TPP’s drug benefit. However, in this case PBM interests are in many instances aligned with the pharmacies in opposition to a principal aspect of the settlements. See note 5, below.
The case is one about drug pricing and some background as to the setting is required.1 The drugs in question are ordi*36narily sold by manufacturers to wholesalers who resell them to pharmacies, who then dispense them to consumers. Where the consumer is insured, customarily the TPP or a PBM reimburses the pharmacy for the drug cost and for a dispensing fee pursuant to a contract with the pharmacy, less any required co-payment made by the consumer directly to the pharmacy.
Drug manufacturers typically sell to drug wholesalers at a list price — called in the industry the “wholesale acquisition cost” (“WAC”) — although discounts may be provided to the wholesaler (e.g., for volume sales). Wholesalers add a mark-up in selling the drugs to retail pharmacies and other purchasers like hospitals. Pharmacies then add a mark-up of their own when they sell the drugs to consumers, some of whom are insured as to their drug purchases and others not.
For drug purchases that are covered by health insurance, the insurer pays for the drugs, apart from any co-payment borne by the consumer; and the consumer is charged only the co-payment. The insurer or its agent typically contracts with the pharmacy to reimburse the latter for the drugs it supplies to the beneficiary based on a discount (which will vary) from a notional benchmark price called the “average wholesale price” (“AWP”). The AWP figure is usually derived by applying a multiplier to the WAC for the drug, and publishers of AWP lists normally obtained their AWP figures from manufacturers or wholesalers. Historically, AWPs were derived by applying different mark-ups to different drugs, the most common multiplier being 1.2 or 1.25 — percentage mark-ups of 20 and 25 percent, respectively.2
Contracts between pharmacies and a TPP or PBM typically incorporate AWP prices by reference. Because the TPP or PBM normally contracts to reimburse pharmacies at a discount from AWP figures, the AWP supplied by a publisher for a drug is likely to be higher than the reimbursement paid by any TPP or PBM; but, given the pharmacy-TPP (or pharmacy-PBM) contract incorporating the AWP as the starting point, an increase in the published AWP means that the TPPs will pay more and the pharmacy will receive more. If the consumer co-payment is a percentage of the drug price, the consumer also pays more. And, to the extent that the pharmacy uses the AWP to set prices for un-insured consumers (see note 2, above), those customers too pay a higher price if the AWP is artificially inflated.
The complaints in this case rest on the premise that McKesson agreed wrongfully with First DataBank to inflate the AWP on over 1,400 drug products from and after August 2001.3 The method used was to inflate the AWP published by First DataBank by using a mark-up of 1.25 over WAC instead of the 1.2 figure that historically had been used for those products. This, in turn, generated higher revenues for pharmacies and higher costs for TPPs and many consumers. This claim, whose merits are not central to this appeal, is *37more fully described elsewhere. New England Carpenters Health Benefits Fund v. First DataBank, Inc., 244 F.R.D. 79 (D.Mass.2007).
The complaint against First DataBank and McKesson was filed on June 2, 2005, and charged a violation of the Racketeer Influenced and Corrupt Organizations Act (“RICO”), 18 U.S.C. § 1962 (2000). Claims were made on behalf of TPPs, possibly some PBMs, consumers whose co-payments were a percentage of a drug price, and un-insured consumers who were charged prices based on AWP figures. How long the harm has continued and how much it has varied over time are less easily determined for reasons that will appear.
In October 2006, a proposed settlement embracing the class claims against First DataBank — but not McKesson — was agreed to by the parties.4 The central provision was that First DataBank agreed to “rollback” its published AWP figures for all drug products (over 8,000 codes) with a mark-up higher than 1.2 down to a 1.2 mark-up. Other provisions included an end to First DataBank’s publishing of AWP within two years, payments of some fees and expenses by the company, and the creation of a data room containing documents for use in other law suits.
The district court granted preliminary approval on November 22, 2006; after amendments were made on May 29, 2007, the court again granted preliminary approval on June 6, 2007. On May 25, 2007, a class action suit was filed against MediSpan for its published AWP list, which was based on First DataBank’s figures, and the parties proposed a settlement the same day. Although the charge was negligence rather than fraud, the settlement was similar to the First DataBank settlement. The district court preliminarily approved the Medi-Span settlement on August 20, 2007.
The proposed settlements prompted objections from most of the challengers now appealing and a few others. The principal objections came from the pharmacy interests, concerned that the rollback would reduce — they said unfairly — the payments to them made by TPPs and PBMs. Three entities — DeVille Pharmacies and two associations representing pharmacies (the Long-Term Care Pharmacy Alliance (“LTCPA”) and the American Society of Consultant Pharmacists (“ASCP"))— moved to intervene; other entities wrote letters to the court and filed comments in opposition.5
The district court held a fairness hearing on the two settlements on January 22, 2008. Athough the hearing was oral, evidence including expert reports was now on file from supporters and opponents of the settlements. After the hearing, the district court rejected the proposed settlements, citing a series of problems. It also rejected as untimely the motions of DeVille Pharmacies, LTCPA and ASCP to intervene, but the objections by the pharmacy interests were nevertheless considered by the court during the hearing.
*38The settling parties filed an amended settlement relating to First DataBank on May 29, 2008, and an amended settlement relating to Medi-Span on July 15, 2008. The centerpiece of these settlements remained the same rollback of AWP earlier proposed, but certain changes were made to address issues raised by the district court,6 which thereafter preliminarily approved the new settlements. Many of the objectors renewed their objections.
The district court then held two more hearings, and issued a decision on March 17, 2009, New England Carpenters Health Benefits Fund v. First DataBank, Inc., 602 F.Supp.2d 277 (D.Mass.2009), approving the settlements. A final order and judgment on March 30, 2009, certified the class and approved the settlements, but delayed implementation of the rollback until September 26, 2009. Although the judgment did not dispose of the claims against McKesson, the district court certified the judgment as final as to First DataBank and Medi-Span, see Fed.R.Civ.P. 54(b), allowing immediate appeal of these settlements.
Various opponents of the settlements— representing pharmacy interests and PBMs — filed appeals that were consolidated for argument, and the case was expedited because of the approach of the rollback date. First DataBank filed a motion to dismiss the appeals, which is pending before us. A number of opponents filed motions, also still pending before us, to intervene in this court. The district court denied a stay pending appeal, and motions to stay in this court were filed but taken under advisement and are disposed of in this decision.
The appeals pivot around three issues: who among the parties seeking to do so, if any, can appear as an appellant in this court; whether the judgment operates against non-party pharmacies so as to violate their due process rights or offend Rule 19, Fed.R.Civ.P. 19, which governs necessary parties; and whether the district court permissibly found that the settlements were “fair, reasonable, and adequate” under Rule 23, Fed.R.Civ.P. 23(e)(2).
Appellant status. That a proper appellant be present is a threshold condition of appellate review and we begin with that issue, sorting the would-be appellants into three groups and addressing each in turn. First, three entities who objected to the settlement in the district court assert status as members of the class: two, the National Association of Chain Drug Stores (“NACDS”) and the Food Marketing Institute (“FMI”), claim that they are themselves TPPs; the last, DeVille Pharmacies, asserts that it is a class member because it purchased drugs at AWP-based prices during the class period. The appellees do not dispute that these three are within the plaintiff class definition.
Class members, unless they opt out, are legally bound by a settlement and so have the easiest claim to status as appellants even though not named as “representative parties” to conduct the case on behalf of unnamed class members. Noting the binding effect of the judgment, Devlin v. Scardelletti, 536 U.S. 1, 10, 122 S.Ct. 2005, 153 L.Ed.2d 27 (2002), held that “nonnamed class members ... who have objected in a timely manner to approval of the settlement at the fairness hearing have *39the power to bring an appeal without first intervening.” Id. at 14, 122 S.Ct. 2005. Thus, a TPP who thought the settlement provided inadequate compensation could readily appeal.
Appellees do not challenge Devlin’s general rule but argue that these three objectors primarily attack the rollback not because compensation for TPPs or other purchasers is inadequate but because it adversely affects the pharmacies’ interest in keeping up their prices. Put more strongly, the objectors’ interests may be to oppose a settlement that serves the interests of TPPs and consumers as a whole but harms objectors who could gain from the settlement as TPPs but lose even more as pharmacies whose prices would fall. This objection raises an unusual set of issues involving the concept of party status, the scope of Devlin and the practical conduct of appellate cases.
Some precedent supports denying appellate status to one who purports to represent another entity or class but whose interests are not aligned or are in opposition.7 But a class member other than a named plaintiff is not a representative; that member is individually bound by a class judgment and is free to pursue his own interest on appeal. Class members even in the same business may be unalike in their interests. Thus, a TPP that had no pharmacy stake might rationally oppose a settlement (e.g., as not rich enough in damages for the TPP) even though risk-averse TPP class members supported it, were right to do so, and were more numerous.
So the real issue is not whether such an atypical member is entitled to appeal but whether — in order to do so — it must first separately intervene in the district court (or in the appeals court). With rare exceptions, a non-party is made to intervene in order to appeal. But a class member is already entitled to appeal under Devlin, unless that case is to be restricted to class members who are also capable of representing class interests. Devlin could be so read since it noted that the would-be appellant’s interests in that case were “within the zone of interests of the requirement” that the settlement be fair to all class members. 536 U.S. at 7,122 S.Ct. 2005.
But Devlin did not decide that an atypical class member had to intervene in order to pursue an appeal, and the quoted language was used in discussing not party status but the separate issue of prudential standing, which is not a concern here. 536 U.S. at 7, 122 S.Ct. 2005. Nor is it clear what useful would be added in this case by requiring intervention when the atypical class member has already objected to the settlement. At the very least, such a condition would require a further inquiry for each Devlin appeal into the class member’s mix of interests — an inquiry solely to decide whether intervention was required to permit an appeal.
Appellees also argue that Devlin does not apply here because appellants had an opportunity to opt out of the settlement class; by contrast, Devlin involved a Rule 23(b)(1) mandatory class action with no opt-out rights. The present class was in fact certified both as a mandatory class and one permitting opt-out rights. In any event, the weight of authority holds that *40Devlin applies to all class actions.8 Devlin, after all, is about party status and one who could cease to be a party is still a party until opting out, which the three entities in question have not done. Other arguments against appellant status were made but require no separate discussion. We conclude that objecting class members have the right to appeal whether their interests are typical or not.
A second set of would-be appellants — ASCP and LTCPA — are not class members but moved to intervene in the district court. They took some months after learning of the proposed settlement in deciding whether to intervene and the district court denied their motions to intervene as untimely. Because prejudice from the delay is not apparent — the same arguments were before the court in the comments they submitted — the refusal to allow intervention might be debatable even though district courts enjoy much latitude on timeliness decisions.
However, these would-be appellants, although entitled to appeal the denials of intervention at once under the collateral order doctrine, Credit Francais Int’l., S.A. v. Bio-Vita, Ltd., 78 F.3d 698, 703 (1st Cir.1996), did not do so and appellees say that it is now too late. Ordinarily, a final judgment brings up all intermediate rulings, but interlocutory appeals of immediately appealable orders must be made within 30 days. Fed. R.App. P. 4(a)(1). As a matter of logic, this might not preclude deferring the appeal until the final judgment but more than one precedent leans against this option when the would-be appellant is not otherwise a party, albeit without detailed consideration.9
By contrast, Ruthardt v. United States, 303 F.3d 375 (1st Cir.2002), cert. denied, 538 U.S. 1031, 123 S.Ct. 2072, 155 L.Ed.2d 1059 (2003), assumed in dicta that (without regard to a possible interlocutory appeal) an appeal from a denial of intervention could occur at the end of the case, id. at 386. Anyway, Ruthardt might have appeared to be First Circuit law when the would-be intervenors forwent immediate appeal. So we here exercise the same option that the court exercised in Ruthardt and, bypassing the question whether intervention was properly denied in the district court, permit it on appeal to those who sought intervention below. Id. Future intervenors should note the yellow warning flag now affixed to the Ruthardt dictum.
A third group of would-be appellants — Eaton Apothecary, Louis and Clark Drug, Thrifty White Drug, the National Community Pharmacists Association, and the Pharmaceutical Care Management Association — neither assert class membership nor attempted to intervene in the district court but claim a right to appeal or appear as parties on appeal simply because their interests are affected by the outcome. The most demonstrable interests are purely practical: that because the decree will lower AWP figures published by the defendants, the pharmacies’ own con*41tracts will cause a reduction in their reimbursement by TPPs and PBMs.
The “general rule” of Marino v. Ortiz, 484 U.S. 301, 108 S.Ct. 586, 98 L.Ed.2d 629 (1988) (per curiam), is that “only parties to a lawsuit, or those that properly become parties, may appeal an adverse judgment.” Id. at 304, 108 S.Ct. 586. Exceptions exist but, as we stressed in Microsystems Software, Inc. v. Scandinavia Online AB, 226 F.3d 35, 40 (1st Cir.2000), they are limited. 15A Wright, Miller & Cooper, Federal Practice and Procedure § 3902.1, at 126-32 (2d ed.1991). Anyway, the fact that a decision against a defendant may practically impact a third party is not ordinarily enough for appellant status absent intervention or joinder in the trial court. Marino, 484 U.S. at 304, 108 S.Ct. 586.
There is no apparent reason here to allow participation as an appellant by one who is neither a class member nor sought to intervene in the district court. The proposed settlements were widely known in the industry, the pharmacy interests are amply represented on appeal and it is hard to see what would be added by further parties. It might be a different matter if the final judgment had purported to alter the legal rights of non-parties, a separate issue to which we now turn.
Legal rights. The pharmacy interests likely do stand to lose revenues, for a limited period, as a result of the rollback intended by the settlement. To the extent that existing reimbursement contracts between pharmacies and TPPs or PBMs are keyed to AWP lists issued by either defendant, a reduction will reduce pharmacy revenue, but both the duration and impact of such reductions are uncertain and almost certainly will vary from one pharmacy to another. Duration and impact issues are pertinent both now and later when we discuss the merits, so some elaboration may be useful.
Duration is uncertain because many pharmacies, TPPs and PBMs renegotiate contracts at regular intervals, some contracts can be renegotiated at will or after trigger events specified in the contracts and possibly (although this is not mentioned) some may be renegotiated because of contract doctrines that apply to unexpected events. See 2 Farnsworth, Farnsworth on Contracts §§ 9.5-9.9 (3d ed.2004). Where rollbacks push reimbursement below market levels, the next contract should remedy that deficiency for future drug purchases — although, where there is a delay in resolution, not without interim loss for the pharmacy.
The character of the impact is also uncertain. The bulk of the inflated AWP figures went into effect in 2001, eight years ago. Seemingly over this period the ordinary processes of bargaining allowed the TPPs and PBMs — many of whom have strength and savvy — to increase the discounts from AWP on which they insist — a process that to some uncertain extent negates the effect of the falsely inflated AWP figures to which the discounts apply. This phenomenon of creeping recapture is the key to the pharmacy claims that the rollback will drive some of them out of business, because a reduction in AWP would only drop pharmacy reimbursements below market levels if recapture had already dissipated some of the previous AWP inflation.10
*42The immediate question is not whether such revenue reductions make the rollback order unfair under Rule 23 standards but whether the rollback violates the Due Process Clause of the Fifth Amendment and Rule 19 of the Federal Rules of Civil Procedure. This dual claim by appellants rests on the fact that almost none of the pharmacy interests were formally parties to the litigation, most of them being not within the class of plaintiffs, nor intervenors in the district court, nor parties involuntarily joined, nor (obviously) defendants.
The constitutional attack itself has two different aspects. The first is a claim that because of the direct financial impact of the rollback, it is unconstitutional to inflict it on one not a party to the case. The axiom is “that a person cannot be deprived of his legal rights in a proceeding to which he is not a party.” Martin v. Wilks, 490 U.S. 755, 758, 109 S.Ct. 2180, 104 L.Ed.2d 835 (1989). Yet the judgment ordering the rollback is not an order that the pharmacies charge less to TPPs or consumers; it is the pharmacies that have chosen by contract to key their reimbursements to figures published by the defendants.
Impact and legal rights are not the same thing. A decision in a contract dispute or antitrust case can have drastic effects on suppliers, stockholders, employees and customers of the company that loses the case; no one thinks the Constitution requires all of them to be parties. Due process “obviously does not mean ... that a court may never issue a judgment that, in practice, affects a nonparty.” Provident Tradesmens Bank & Trust Co. v. Patterson, 390 U.S. 102, 110, 88 S.Ct. 733, 19 L.Ed.2d 936 (1968); see also James & Hazard, Civil Procedure § 10.12, at 534 (3d ed.1985).
Instead, impacted non-parties can seek to intervene or otherwise express their views in litigation that may affect their practical interests — the fairness hearing required by Rule 23(e)(2) provides just such a mechanism. Fed.R.Civ.P. 23(e)(2); In re Masters Mates & Pilots Pension Plan & IRAP Litig., 957 F.2d 1020, 1031 (2d Cir.1992). In rare instances, as provided by Rule 19(b), a court may be empowered not to decide litigation between the parties because non-party interests cannot be fairly protected. Fed. R.Civ.P. 19(b). But in general, the opportunity for non-parties to protect their practical interests is considerable and sufficient.
The pharmacy interests have a second string to their constitutional bow: they say that this judgment was accompanied by findings or characterizations by the district court that would control in future litigation against them and create legal liability for them in that litigation. Principally, they point to the district court’s comments in colloquy and the final decision that the pharmacies objecting to the rollback were themselves “unjustly enriched” by the inflated AWP prices or received a “windfall.”
Unjust enrichment, in a legal framework, comprises a claim or claims on which relief may be granted in a lawsuit by the person unjustly deprived.11 In this case, the district court did not decide unjust enrichment claims: it was saying as a *43matter of fact, which can hardly be denied, that unexpected inflation in AWP rates would result in many pharmacies getting unexpected extra revenues which a later downward adjustment would reverse. This is a lay use of the terms made in assessing whether the settlements were reasonable under Rule 23.
Class action counsel in this case are now arguing to at least one other court that the district court here found that the pharmacies were “unjustly enriched.” Class Action Complaint ¶ 3, Skilstaf, Inc. v. CVS Caremark Corp., Case No. 3:09-cv-02514-SI, 2009 WL 2749909 (N.D. Cal. filed June 5, 2009). Even ignoring limits on res judicata that might be invoked, this is to confound an epithet with a legal ruling on a claim not before the district judge, and a sister court is unlikely to be fooled by the argument.
The other legal-rights argument made to us rests on Rule 19(a)(l)(B)(i), providing that a person who could be made a party to a case must, under certain circumstances, be joined; pertinently, if the “person claims an interest relating to the subject of the action” and resolution without that person may “as a practical matter impair or impede the person’s ability to protect the interest.” Fed.R.Civ.P. 19(a)(l)(B)(i). If the person cannot feasibly be joined, “the court must determine whether, in equity and good conscience,” the case should proceed without him. Fed.R.Civ.P. 19(b).
No one in the district court, including the pharmacies and PBMs who were well represented by their associations, suggested to the district court that Rule 19 was being thwarted. That a rollback might be ordered was made known early on in the settlement approval process, and opposing comments by the pharmacy interests were extensive. Appellants now say that the unjust enrichment and windfall comments came only in the final decision and so too late to protest, but these terms were used in oral hearings well before this.12 In addition, a Rule 19 issue arising only because of comments in the final decision could easily have been called to the district court’s attention afterwards.
So for themselves the appellants have forfeited the issue, subject to claims of plain error. Successful plain error claims are rare in civil cases, Chestnut v. City of Lowell, 305 F.3d 18, 20 (1st Cir.2002) (en banc), and impossible here, given the loose phrasing of the Rule 19 criteria. Courts have some independent duty to protect absent “necessary” parties, Provident, 390 U.S. at 111, 88 S.Ct. 733, from any threat of serious injustice, and we could invoke Rule 19 sua sponte. But for two different reasons no such measure is appropriate.
First, it is far from clear that there is any violation of Rule 19; if the absent pharmacies and PBMs were deemed necessary, most could not practically be joined in this or any other law suit because of their numerosity. And, in addition, Rule 19 dismissals are rarely appropriate when the objection is first made at the end of the case. See Provident, 390 U.S. at 109-10, 88 S.Ct. 733; GTE Sylvania, Inc. v. Consumer Prod. Safety Comm’n, 598 F.2d 790, 798 (3d. Cir.1979), aff'd 447 U.S. 102, 100 S.Ct. 2051, 64 L.Ed.2d 766 (1980). Second, the interests of the absent pharmacies and PBMs have been vigorously addressed by arguments *44and evidence from pharmacy interests who were and are present and were considered by the district court in passing on the settlements.
The merits of the settlements. The district judge must approve the settlement in a class action and, to do so, must allow a hearing and make a finding that the settlement “is fair, reasonable, and adequate,” Fed.R.Civ.P. 23(e)(2), or (in shorthand), “reasonable.” In this case, there were successive hearings, numerous submissions by the parties including expert reports, several decisions on different proposals and a final decision on March 17, 2009, evaluating and approving the final settlements. The final judgment contains the formal findings required by Rule 23.
Rule 23’s reasonableness standard has been given substance by case law offering laundry lists of factors, most of them intuitively obvious and dependent largely on variables that are hard to quantify; usually, the ultimate decision by the judge involves balancing the advantages and disadvantages of the proposed settlement as against the consequences of going to trial or other possible but perhaps unattainable variations on the proffered settlement.13
Here, the pharmacy interests’ attack is — with one qualification — not about the articulation of the standard but its alleged misapplication to the facts of this case. The principal misapplication claim, to be described shortly in more detail, is that the rollback of AWP figures by the defendants will seriously (and unfairly) curtail pharmacy income and may drive a considerable number of pharmacies out of business. In assessing the attack (and a few less central objections considered thereafter), three legal propositions govern our review.
First, the interests of non-parties to the settlement must be taken into account: “[I]f third parties will be affected, [the court must find that the settlement] will not be unreasonable or legally impermissible as to them.” Durrett v. Housing Auth. of City of Providence, 896 F.2d 600, 604 (1st Cir.1990); see also In re Masters Mates, 957 F.2d at 1026. In general, affected non-parties may make submissions or seek to intervene; here, the principal opposition was in fact from the pharmacy interests and PBM representatives who are predominantly non-parties. But the interests of non-parties are just one element in the equation.14
Second, although policy encourages settlements, Durrett, 896 F.2d at 604, the burden remains on the proponents to show that the settlement is reasonable, Greenspun v. Bogan, 492 F.2d 375, 378 (1st Cir.1974). Usually, “there is a presumption in favor of the settlement” if discovery has been adequate and the parties have bargained at arms length, City P’ship Co. v. Atl. Acquisition Ltd. P’ship, 100 F.3d 1041, 1043 (1st Cir.1996), but we think that presumption is inapposite in judging the impact on non-parties who were not represented in the bargaining.
*45Third, the district court enjoys considerable range in approving or disapproving a class action settlement, given the generality of the standard and the need to balance benefits and costs. The usual rubric for appellate review is abuse of discretion. City P’ship Co., 100 F.3d at 1043-44. This over-simplifies: embedded legal issues are reviewed de novo, see, e.g., Durrett, 896 F.2d at 603; Cent. States Se. & Sw. Areas Health & Welfare Fund v. Merck-Medco Managed Care, L.L.C., 504 F.3d 229, 247 (2d Cir.2007), and factual findings for clear error, Fed.R.Civ.P. 52(a)(6). But the latter is also a deferential standard, and here the significant legal issues — concerning due process and Rule 19 — have already been dispatched.
The arguments on both sides are easily condensed. The pharmacy interests claim that they will suffer severe financial harms including business failures for pharmacies and renegotiation costs to PBMs; that any benefit the pharmacies got was innocent and inflation in net prices has been eroded by renegotiation since the original increase; that some pharmacies will receive unduly low prices for at least some period; and that the settling defendants who are parties to these appeals will pay only a modest amount of money (roughly $2.7 million) while the pharmacies who committed no fraud will bear the brunt of the burden of the settlement. The principal evidence in support is contained in expert reports.15
Conversely, the defenders of the rollback say that the AWP figures were wrongfully inflated and undoing that inflation is a step toward precluding future harm from continued over-pricing; that the pharmacies were enriched and the rollback will give only a portion back to payors; that even if TPP reimbursement levels have been bargained down to some extent, some lingering inflation remains to be undone; that if the rollback forces some reimbursement levels too low, contract changes between TPPs, PBMs and pharmacies will readjust the figures over time; and that significant pharmacy failures, beyond consolidation that is occurring anyway in the industry, are speculative.16
In principle, the rollback makes some sense: it should — to the extent that prices remain above some hypothetical market level — wash out any remaining inflation for the future; and, to the extent it forces prices temporarily below the market level, it will take back some of the windfall profits obtained (even if innocently through another’s fraud) and give some compensation for past overcharges. The market forces that eroded excess gains in the past should also in the future redress unduly low reimbursement; and to the extent that those forces operate symmetrically, the gains and the losses may even tend to balance out.17
*46But many variables affect the outcome and reliable figures are hard to come by — • inevitably so as to future effects which depend on how numerous parties react, and how quickly, to any settlement. As a baseline, a proponent expert estimated that the inflation of AWP has cost purchasers $7 billion and the rollback might save as much as $1 billion in the first year in payments (subject to erosion thereafter), a figure opponents say is high. But all that seems predictable is a substantial temporary benefit to the class, eroding over time, and likely to be much less than the original losses.
The impact on pharmacies will depend on several of these factors about which the experts argue. One expert for the pharmacy interests claimed that 40 percent of existing non-chain community pharmacies will be driven from the market or threatened with this fate. But the underlying calculation is dubious and the result intuitively far fetched, and an expert for the proponents of the settlement estimated a much smaller revenue effect.18 Further, the TPPs have a substantial interest in not choking off their distribution channels and enhancing the power of those who remain. To the extent benefits are overestimated, so too is adverse impact.
One might argue that an inability to be sure as to consequences ought to doom the rollback. Yet presumptively the rollback makes sense absent extreme circumstances — say, gains for buyers greatly exceeding the original losses or likely to force large numbers of pharmacies out of business. Such adverse effects, one might think, are something that the opponents of the rollback ought to establish. They certainly have not done so. As the record stands, the rollback — taken by itself — was within the ambit of reasonable choices available to the district judge.
Looking at the remedy in relation to what the defendants are contributing is a separate matter. Arguably, the very best solution would be to make the settling defendants shoulder the full cost. But whether or not the publisher defendants could pay a bit more than $2.7 million, it would not make even a small dent in a multi-billion dollar loss. McKesson — not a party to these appeals- — agreed in its settlement (which the district court approved) to pay $350 million, but this is far short of the total estimated damage. Opponents of the settlement have not shown that the defendants collectively could have been made to pay more.
The remaining objections appear to be small beer. Some challengers claim that certain drug codes covered in the rollback have been discontinued or were included in the second proposed settlement but not the first. After receiving expert reports and holding a hearing on this issue, the district court rejected these objections for what appear to be good reasons, namely, harm from inclusion was not apparent, discontinued drug codes might be re-introduced, and the other drug codes were properly included in the settlement.
*47The challengers also object to the settlements including the creation of “data rooms” with information and documents from First DataBank and Medi-Span on the grounds that these rooms will include proprietary TPP, pharmacy and PBM information, and that the rooms will not benefit the class. The district court found that the rooms would not include proprietary information, a finding not shown to be clearly erroneous. Nor does the provision exceed the court’s authority: the data rooms may well benefit the class by streamlining discovery in related law suits while protecting the defendants against duplicative discovery.
The district court’s management of the ease, insistence on a revised settlement and multiple hearings have given the pharmacy interests a lengthy period to prepare for the rollback — a period extended by the district judge to 180 days after the final judgment. We affirm, the final judgment and deny pending motions for a stay of the final judgment. Motions to intervene in this court by ASCP and LTCPA are granted and all other motions to intervene are denied. We deny the pending motion to dismiss with respect to the parties that we find can appeal or intervene — namely, NACDS, FMI, DeVille Pharmacies, ASCP and LTCPA — and grant the motion to dismiss as to the other parties.
It is so ordered.