POSNER, Chief Judge.
Drug companies that manufacture blood solids are the defendants in a nationwide class action brought on behalf of hemophiliacs infected by the AIDS virus as a consequence of using the defendants’ products. The defendants have filed with us a petition for mandamus, asking us to direct the district judge to rescind his order certifying the case as a class action. We have no appellate jurisdiction over that order. An order certifying a class is not a final decision within the meaning of 28 U.S.C. § 1291; it does not wind up the litigation in the district court. And, in part because it is reviewable (at least in principle — the importance of this qualification will appear shortly) on appeal from the final decision in the case, it has been held not to fit any of the exceptions to the rule that confines federal appellate jurisdiction to final decisions. In short, as the Supreme Court made clear in Coopers & Lybrand v. Livesay, 437 U.S. 463, 98 S.Ct. 2454, 57 L.Ed.2d 351 (1978), and Gardner v. Westinghouse Broadcasting Co., 437 U.S. 478, 480-82, 98 S.Ct. 2451, 2453-54, 57 L.Ed.2d 364 (1978), it is not an appealable order. Those decisions involved the denial rather than the grant of motions for class certification, but the grant is no more final than the denial and no more within any of the exceptions to the final-decision rule. Hoxworth v. Blinder, Robinson & Co., 903 F.2d 186, 208 (3d Cir.1990); 7B Charles Alan Wright, Arthur R. Miller & Mary Kay Kane, Federal Practice and Procedure § 1802, pp. 484-86 (2d ed. 1986). Still, even nonappealable orders can be challenged by asking the court of appeals to mandamus the district court. Indeed, as a practical matter only such orders can be challenged by filing a petition for mandamus; an appealable order can be challenged only by appealing from it; the possibility of appealing would be a compelling reason for denying mandamus. For obvious reasons, however, mandamus is issued only in extraordinary cases. Otherwise, interlocutory orders would be appealable routinely, but with “appeal” renamed “mandamus.” Kerr v. United States District Court, 426 U.S. 394, 403, 96 S.Ct. 2119, 2124, 48. L.Ed.2d 725 (1976); Eisenberg v. United [1295]*1295States District Court, 910 F.2d 374, 375 (7th Cir.1990).
How to cabin this too-powerful writ which if uncabined threatens to unravel the final-decision rule? By taking seriously the two conditions for the grant of a writ of mandamus. The first is that the challenged order not be effectively reviewable at the end of the case—in other words, that it inflict irreparable harm. Kerr v. United States, supra, 426 U.S. at 403, 96 S.Ct. at 2124; In re Sandahl, 980 F.2d 1118, 1119 (7th Cir.1992); Eisenberg v. United States District Court, supra, 910 F.2d at 375. The petitioner “must ordinarily demonstrate that something about the order, or its circumstances, would make an end-of-case appeal ineffectual or leave legitimate interests unduly at risk.” In re Recticel Foam Corp., 859 F.2d 1000, 1005-06 (1st Cir.1988). Second, the order must so far exceed the proper bounds of judicial discretion as to be legitimately considered usurpative in character, or in violation of a clear and indisputable legal right, or, at the very least, patently erroneous. Gulfstream Aerospace Corp. v. Mayacamas Corp., 485 U.S. 271, 289, 108 S.Ct. 1133, 1143-44, 99 L.Ed.2d 296 (1988); Allied Chemical Corp. v. Daiflon, Inc., 449 U.S. 33, 35, 101 S.Ct. 188, 190, 66 L.Ed.2d 193 (1980) (per curiam); United States v. Spilotro, 884 F.2d 1003, 1006-07 (7th Cir.1989); In re Sandahl, supra, 980 F.2d at 1121; Maloney v. Plunkett, 854 F.2d 152 (7th Cir.1988). We shall not have to explore these gradations; it will be enough to consider whether the district judge’s order can fairly be characterized as usurpative.
The set of orders in which both conditions are satisfied is small. It certainly is not coterminous with the set of orders certifying suits as class actions. For even though such orders often, perhaps typically, inflict irreparable injury on the defendants (just as orders denying class certification often, perhaps typically, inflict irreparable injury on the members of the class), irreparable injury is not sufficient for mandamus; there must also be an abuse of discretion that can fairly be characterized as gross, very clear, or unusually serious. But it is not an empty set. The point of cases like Coopers & Lybrand is that irreparable harm is not enough to make class certification orders automatically ap-pealable under 28 U.S.C. § 1291, not that mandamus is never appropriate in a class certification setting. There is a big difference between saying that all class certification rulings are appealable as of right because they are final within the meaning of section 1291 (the position rejected in Coopers & Lybrand) and saying that a handful are— the handful in which the district judge committed a clear abuse of discretion. Mandamus has occasionally been granted to undo class certifications, see, e.g., In re Fibreboard Corp., 893 F.2d 706 (5th Cir.1990), and we are not aware that any case has held that mandamus will never be granted in such cases. See In re Catawba Indian Tribe, 973 F.2d 1133, 1137 (4th Cir.1992); DeMasi v. Weiss, 669 F.2d 114, 117-19 and n. 6 (3d Cir.1982). The present case, as we shall see, is quite extraordinary when all its dimensions are apprehended. We shall also see that when mandamus is sought to protect the Seventh Amendment’s right to a jury trial in federal civil cases, as in this case, the requirement of proving irreparable harm is relaxed.
The suit to which the petition for mandamus relates, Wadleigh v. Rhone-Poulenc Rarer Inc., 157 F.R.D. 410 arises out of the infection of a substantial fraction of the hemophiliac population of this country by the AIDS virus because the blood supply was contaminated by the virus before the nature of the disease was well understood or adequate methods of screening the blood supply existed. The AIDS virus (HIV—human immunodeficiency virus) is transmitted by the exchange of bodily fluids, primarily semen and blood. Hemophiliacs depend on blood solids that contain the clotting factors whose absence .defines their disease. These blood solids are concentrated from blood obtained from many donors. If just one of the donors is infected with the AIDS virus the probability that the blood solids manufactured in part from his blood will be infected is very high unless the blood is treated with heat to kill the virus. For general background, see Margaret W. Hilgartner, “AIDS and Hemophilia,” 317 New England Journal of Medicine 1153 (1987); Leon W. Hoyer, “Hemophilia [1296]*1296A,” 330 New England Journal of Medicine 38 (1994); “U.S. CDC: HIV Cutting Lives Short in Hemophilia, Study Says,” AIDS Weekly, Feb. 14, 1994.
First identified in 1981, AIDS was diagnosed in hemophiliacs beginning in 1982, and by 1984 the medical community agreed that the virus was transmitted by blood as well as by semen. That year it was demonstrated that treatment with heat could kill the virus in the blood supply and in the following year a reliable test for the presence of the virus in blood was developed. By this time, however, a large number of hemophiliacs had become infected. Since 1984 physicians have been advised to place hemophiliacs on heat-treated blood solids, and since 1985 all blood donated for the manufacture of blood solids has been screened and supplies discovered to be HIV-positive have been discarded. Supplies that test negative still are heat-treated, because the test is not infallible and in particular may fail to detect the virus in persons who became infected within six months before taking the test.
The plaintiffs have presented evidence that 2,000 hemophiliacs have died of AIDS and that half or more of the remaining U.S. hemophiliac population of 20,000 may be HIV-positive. Unless there are dramatic breakthroughs in the treatment of HIV or AIDS, all infected persons will die from the disease. The reason so many are infected even though the supply of blood for the manufacture of blood solids (as for transfusions) has been safe since the mid-80s is that the disease has a very long incubation period; the median period for hemophiliacs may be as long as 11 years. Probably most of the hemophiliacs who are now HIV-positive, or have AIDS, or have died of AIDS were infected in the early 1980s, when the blood supply was contaminated.
Some 300 lawsuits, involving some 400 plaintiffs, have been filed, 60 percent of them in state courts, 40 percent in federal district courts under the diversity jurisdiction, seeking to impose tort liability on the defendants for the transmission of HIV to hemophiliacs in blood solids manufactured by the defendants. Obviously these 400 plaintiffs represent only a small fraction of the hemophiliacs (or their next of kin, in cases in which the hemophiliac has died) who are infected by HIV or have died of AIDS. One of the 300 cases is Wadleigh, filed in September 1993, the case that the district judge certified as a class action. Thirteen other cases have been tried already in various courts around the country, and the defendants have won twelve of them. All the cases brought in federal court (like Wadleigh) — cases brought under the diversity jurisdiction — have been consolidated for pretrial discovery in the Northern District of Illinois by the panel on multidis-trict litigation.
The plaintiffs advance two principal theories of liability. The first is that before anyone had heard of AIDS or HIV, it was known that Hepatitis B, a lethal disease though less so than HIV-AIDS, could be transmitted either through blood transfusions or through injection of blood solids. The plaintiffs argue that due care with respect to the risk of infection with Hepatitis B required the defendants to take measures to purge that virus from their blood solids, whether by treating the blood they bought or by screening the donors — perhaps by refusing to deal with paid donors, known to be a class at high risk of being infected with Hepatitis B. The defendants’ failure to take effective measures was, the plaintiffs claim, negligent. Had the defendants not been negligent, the plaintiffs further argue, hemophiliacs would have been protected not only against Hepatitis B but also, albeit fortuitously or as the plaintiffs put it “serendipitously,” against HIV.
The plaintiffs’ second theory of liability is more conventional. It is that the defendants, again negligently, dragged their heels in screening donors and taking other measures to prevent contamination of blood solids by HIV when they learned about the disease in the early 1980s. The plaintiffs have other theories of liability as well, including strict products liability, but it is not necessary for us to get into them.
The district judge did not think it feasible to certify Wadleigh as a class action for the adjudication of the entire controversy between the plaintiffs and the defendants. Fed.R.Civ.P. 23(b)(3). The differences in the [1297]*1297date of infection alone of the thousands of potential class members would make such a procedure infeasible. Hemophiliacs infected before anyone knew about the contamination of blood solids by HIV could not rely on the second theory of liability, while hemophiliacs infected after the blood supply became safe (not perfectly safe, but nearly so) probably were not infected by any of the defendants’ products. Instead the judge certified the suit “as a class action with respect to particular issues” only. Fed.R.Civ.P. 23(e)(4)(A). He explained this decision in an opinion which implied that he did not envisage the entry of a final judgment but rather the rendition by a jury of a special verdict that would answer a number of questions bearing, perhaps decisively, on whether the defen-, dants are negligent under either of the theories sketched above. If the special verdict found no negligence under either theory, that presumably would be the end of all the cases unless other theories of liability proved viable. If the special verdict found negligence, individual members of the class would then file individual tort suits in state and federal district courts around the nation and would use the special verdict, in conjunction with the doctrine of collateral estoppel, to block relitigation of the issue of negligence.
With all due respect for the district judge’s commendable desire to experiment with an innovative procedure for streamlining the adjudication of this “mass tort,” we believe that his plan so far. exceeds the permissible bounds of discretion in the management of federal litigation as to compel us to intervene and order decertification. The plaintiffs’ able counsel argues that we need not intervene now, that it will be time enough to intervene if and when a special verdict adverse to the defendants is entered and an appeal taken to us. But of course a verdict as such is not an appealable order. Only when a final judgment is entered, determining liability and assessing damages, will the case, including interim rulings such as the certification of certain issues in the case for determination in a class action, be appealable to us. Since without a final judgment the special verdict would not (with an exception noted later in this opinion) even have collateral estoppel effect, Amcast Industrial Corp. v. Detrex Corp., 45 F.3d 155, 158 (7th Cir.1995), the district judge may have intended that the special verdict would be followed by a trial on any remaining liability issues, and on damages, limited to Wadleigh and the other named plaintiffs in the Wadleigh case. That trial would culminate in a final judgment, which would both be appealable to us and impart collateral estoppel effect to the special verdict. The members of the class, other than the named plaintiffs, would take the special verdict back to their home districts and use it to limit the scope of the individual trials that would be necessary — for remember that the district judge has refused to certify the case as a class action for a final adjudication of the controversy between the class and the defendants — to determine each class member’s actual entitlement to damages and in what amount.
We asked at argument whether the district judge indeed planned to enter a final judgment at least with respect to the named plaintiffs, so that the ruling on class certification could eventually be appealed. The plaintiffs’ counsel relayed the question to the judge, who wrote them that he does envisage the eventual entry of a final judgment with regard to the named plaintiffs. This procedure for eliciting the judge’s views was irregular. A judge against whom mandamus is sought is authorized to file a' brief, Fed. R.App.P. 2Í(b), and Judge Grady had not done so. In effect the letter was his brief, but because it was filed out of time the petitioners had no chance to respond to it.
Nevertheless we shall assume in accordance with the judge’s letter that eventually there will be a final judgment to review. Only it will come too late to provide effective relief to the defendants; and this is an important consideration in relation to the first condition for mandamus, that the challenged ruling of the district court have inflicted irreparable harm, which is to say harm that cannot be rectified by an appeal from the final judgment in the lawsuit. The reason that an, appeal will come too late to provide effective relief for these defendants is the sheer magnitude of the risk to which the class action, in contrast to the individual actions pending or likely, exposes them. [1298]*1298Consider the situation that would obtain if the class had not been certified. The defendants would be facing 300 suits. More might be filed, but probably only a few more, because the statutes of limitations in the various states are rapidly expiring for potential plaintiffs. The blood supply has been safe since 1985. That is ten years ago. The risk to hemophiliacs of having become infected with HIV has been widely publicized; it is unlikely that many hemophiliacs are unaware of it. Under the usual discovery statute of limitations, they would have to have taken steps years ago to determine their infection status, and having found out file suit within the limitations period running from the date of discovery, in order to preserve their rights.
Three hundred is not a trivial number of lawsuits. The potential damages in each one are great. But the defendants have won twelve of the first thirteen, and, if this is a representative sample, they are likely to win most of the remaining ones as well. Perhaps in the end, if class-action treatment is denied (it has beén denied in all the other hemophiliac HIV suits in which class certification has been sought), they will be compelled to pay damages in only 25 cases, involving a potential liability of perhaps no more than $125 million altogether. These are guesses, of course, but they are at once conservative and usable for the limited purpose of comparing the situation that will face the defendants if the class certification stands. All of a sudden they will face thousands of plaintiffs. Many may already be barred by the statute of limitations, as we have suggested, though its further running was tolled by the filing of Wadleigh as a class action. American Pipe & Construction Co. v. Utah, 414 U.S. 538, 554, 94 S.Ct. 756, 766, 38 L.Ed.2d 713 (1974). (If the class is decertified, the statute of limitations will start running again. Glidden v. Chromalloy American Corp., 808 F.2d 621, 627 (7th Cir.1986); Barrett v. U.S. Civil Service Comm’n, 439 F.Supp. 216, 218-19 (D.D.C.1977); cf. American Pipe & Construction Co. v. Utah, supra, 414 U.S. at 552, 561, 94 S.Ct. at 765, 770; Crown, Cork & Seal Co. v. Parker, 462 U.S. 345, 354, 103 S.Ct. 2392, 2398, 76 L.Ed.2d 628 (1983)). Suppose that 5,000 of the potential class members are not yet barred by the statute of limitations. And suppose the named plaintiffs in Wadleigh win the class portion of this ease to the extent of establishing the defendants’ liability under either of the two negligence theories. It is true that this would only be prima facie liability, that the defendants would have various defenses. But they could not be confident that the defenses would prevail. They might, therefore, easily be facing $25 billion in potential liability (conceivably more), and with it bankruptcy. They may not wish to roll these dice. That is putting it mildly. They will be under intense pressure to settle. Milton Handler, “The Shift from Substantive to Procedural Innovations in Antitrust Suits,” 71 Column.L.Rev. 1, 8-9 (1971); William Simon, “Class Actions — Useful Tool or Engine of Destruction,” 55 F.R.D. 375 (1972); Marc Galanter, “Why the ‘Haves’ Come Out Ahead: Speculations on the Limits of Legal Change,” 9 Law & Soc’y Rev. 95, 143 and n. 121 (1974); Charles D. Schoor, “Class Actions: The Right to Solicit,” 16 Santa Clara L.Rev. 215, 239-40 and n. 82 (1976); Joseph Grundfest, “Disimplying Private Rights of Action under the Federal Securities Laws: The Commission’s Authority,” 107 Harv. L.Rev. 963, 973 n. 38 (1994); Lester Brick-man, “On the Relevance of the Admissibility of Scientific Evidence: Tort System Outcomes Are Principally Determined by Lawyer’s Rates of Return,” 15 Cardozo L.Rev. 1755, 1780-82 (1994); Note, “Conflicts in Class Actions and Protection of Absent Class Members,” 91 Yale L.J. 590, 605 n. 67 (1982). If they settle, the class certification — the ruling that will have forced them to settle — will never be reviewed. General Motors Corp. v. City of New York, 501 F.2d 639, 657-58 (2d Cir.1974) (concurring opinion); cf. Mars Steel Corp. v. Continental Illinois National Bank & Trust Co., 834 F.2d 677, 682-83 (7th Cir.1987). Judge Friendly, who was not given to hyperbole, called settlements induced by a small probability of an immense judgment in a class action “blackmail settlements.” Henry J. Friendly, Federal Jurisdiction: A General View 120 (1973). Judicial concern about them is legitimate, not “sociological,” as it was derisively termed in In re Sugar Anti[1299]*1299trust Litigation, 559 F.2d 481, 483 n. 1 (9th Cir.1977).
The defendants did not mention their concern about settlement pressures until the oral argument of this appeal, so we should consider whether the argument is waived and we therefore cannot consider it. If a party fails to present a ground for reversal, the appeals court will not supply it; this is the doctrine of waiver. Hartmann v. Prudential Ins. Co., 9 F.3d 1207, 1214 (7th Cir.1993); United States v. Rodriguez, 888 F.2d 519, 524 (7th Cir.1989); Bonds v. Coca-Cola Co., 806 F.2d 1324, 1328 (7th Cir.1986). The doctrine is not that if the party fails to offer a particular reason for its position, the court cannot consider that reason. Transcraft, Inc. v. Galvin, Stalmack, Kirschner & Clark, 39 F.3d 812, 820 (7th Cir.1994). Were that the rule, the role of an appellate court would be confined to weighing the reasons, pro and con a particular ground, that the parties happened to proffer. Appellate consideration so truncated could not produce durable rules to guide decision in future cases; judicial opinions would be impoverished if all they did were call balls and strikes. The defendants here properly asserted as the basis for mandamus the two necessary conditions: irreparable harm and clear violation of right. For obvious reasons they did not point out in support of the first condition that if mandamus is denied they will be forced to settle — for such an acknowledgment would greatly weaken them in any settlement negotiations. We should be realistic about what is feasible to put in a public brief.
We do not want to be misunderstood as saying that class actions are bad because they place pressure on defendants to settle. That pressure is a reality, but it must be balanced against the undoubted benefits of the class action that have made it an authorized procedure for employment by federal courts. We have yet to consider the balance. All that our discussion to this point has shown is that the first condition for the grant of mandamus — that the challenged ruling not be effectively reviewable at the end of the case — is fulfilled. The ruling will inflict irreparable harm; the next question is whether the ruling can fairly be described as usur-pative. We have formulated this second condition as narrowly, as stringently, as can be, but even so formulated we think it is fulfilled. We do not mean to suggest that the district judge is engaged in a deliberate power-grab. We have no reason to suppose that he wants to preside over an unwieldy class action. We believe that he was responding imaginatively and in the best of faith to the challenge that mass torts, graphically illustrated by the avalanche of asbestos litigation, pose for the federal courts. But the plan that he has devised for the HIV-hemophilia litigation exceeds the bounds of allowable judicial discretion. Three concerns, none of them necessarily sufficient in itself but cumulatively compelling, persuade us to this conclusion.
The first is a concern with forcing these defendants to stake their companies on the outcome of a single jury trial, or be forced by fear of the risk of bankruptcy to settle even if they have no legal liability, when it is entirely feasible to allow a final, authoritative . determination of their liability for the colossal misfortune that has befallen the hemophiliac population to emerge from a decentralized process of multiple trials, involving different juries, and different standards of liability, in different jurisdictions; and when, in addition, the preliminary indications are that the defendants are not liable for the grievous harm that has befallen the members of the class. These qualifications are important. In most class actions — and those the ones in which the rationale for the procedure is most compelling — individual suits are infeasible because the claim of each class member is tiny relative to the expense of litigation. That plainly is not the situation here. A notable feature of this case, and one that has not been remarked upon or encountered, so far as we are aware, in previous eases, is the demonstrated great likelihood that the plaintiffs’ claims, despite their human appeal, lack legal merit. This is the inference from the defendants’ having won 92.3 percent (12/13) of the cases to have gone to - judgment. Granted, thirteen is a small sample and further trials, if they are held, may alter the pattern that the sample reveals. But whether they do or not, the result will be robust if these further trials are permitted to go forward, because the pattern that results will [1300]*1300reflect a consensus, or at least a pooling of judgment, of many different tribunals.
For this consensus or maturing of judgment the district judge proposes to substitute a single trial before a single jury instructed in accordance with no actual law of any jurisdiction — a jury that will receive a kind of Esperanto instruction, merging the negligence standards of the 50 states and the District of Columbia. One jury, consisting of six persons (the standard federal civil jury nowadays consists of six regular jurors and two alternates), will hold the fate of an industry in the palm of its hand. This jury, jury number fourteen, may disagree with twelve of the previous thirteen juries — and hurl the industry into bankruptcy. That kind of thing can happen in our system of civil justice (it is not likely to happen, because the industry is likely to settle — whether or not it really is liable) without violating anyone’s legal rights. But it need not be tolerated when the alternative exists of submitting an issue to multiple juries constituting in the aggregate a' much larger and more diverse sample of decision-makers. That would not be a feasible option if the stakes to each class member were too slight to repay the cost of suit, even though the aggregate stakes were very large and would repay the costs of a consolidated proceeding. But this is not the case with regard to the HIV-hemophilia litigation. Each plaintiff if successful is apt to receive a judgment in the millions. With the aggregate stakes in the tens or hundreds of millions of dollars, or even in the billions, it is not a waste of judicial resources to conduct more than one trial, before more than six jurors, to determine whether a major segment of the international pharmaceutical industry is to follow the asbestos manufacturers into Chapter 11.
We have hinted at the second reason for concern that the district judge exceeded the bounds of permissible judicial discretion. He proposes to have a jury determine the negligence of the defendants under a legal standard that does not actually exist anywhere in the world. One is put in mind of the concept of “general” common law that prevailed in the era of Swift v. Tyson. The assumption is that the common law of the 50 states and the District of Columbia, at least so far as bears on a claim of negligence against drug companies, is basically uniform and can be abstracted in a single instruction. It is no doubt true that at some level of generality the law of negligence is one, not only nationwide but worldwide. Negligence is a failure to take due care, and due care a function of the probability and magnitude of an accident and the costs of avoiding it. A jury can be asked whether the defendants took due care. And in many cases such differences as there are among the tort rules of the different states would not affect the outcome. The Second Circuit was willing to assume dubitante that this was true of the issues certified for class determination in the Agent Orange litigation. In re Diamond Shamrock Chemicals Co., 725 F.2d 858, 861 (2d Cir.1984).
We doubt that it is true in general, and we greatly doubt that it is true in a case such as this in which one of the theories pressed by the plaintiffs, the “serendipity” theory, is novel. If one instruction on negligence will serve to instruct the jury on the legal standard of every state of the United States applicable to a novel claim, implying that the claim despite its controversiality would be decided identically in all 50 states and the District of Columbia, one wonders what the Supreme Court thought it was doing in the Erie case when it held that it was unconstitutional for federal courts in diversity cases to apply general common law rather than the common law of the state whose law would apply if the ease were -being tried in state rather than federal court. Erie R.R. v. Tompkins, 304 U.S. 64, 78-80, 58 S.Ct. 817, 822, 82 L.Ed. 1188 (1938). The law of negligence, including subsidiary concepts such as duty of care, foreseeability, and proximate cause, may as the plaintiffs have argued forcefully to us differ among the states only in nuance, though we think not, for a reason discussed later. But nuance can be important, and its significance is suggested by a comparison of differing state pattern instructions on negligence and differing judicial formulations of the meaning of negligence and the subordinate concepts. See, e.g., McCarty v. Pheasant Run, Inc., 826 F.2d 1554, 1556-57 (7th Cir.1987); Olson v. Prosoco, Inc., 522 N.W.2d 284, 286-90 (Ia.1994); Ar [1301]*1301kansas Kraft v. Cottrell, 313 Ark. 465, 855 S.W.2d 333, 337 (1993); Flight Line, Inc. v. Tanksley, 608 So.2d 1149, 1158-59 (Miss.1992); Knight v. Jewett, 3 Cal. 4th 296, 11 Cal.Rptr.2d 2, 834 P.2d 696 (1992); Doe v. Grosvenor Properties (Hawaii) Ltd., 73 Haw. 158, 829 P.2d 512 (1992); Riddle v. McLouth Steel Products Corp., 440 Mich. 85, 485 N.W.2d 676 (1992); Kinsey v. Bray, 596 N.E.2d 938, 940 (Ind.App.1992); Preston v. Keith, 217 Conn. 12, 584 A.2d 439 (1991); Kalata v. Anheuser-Busch Cos., 144 Ill.2d 425, 163 Ill.Dec. 502, 581 N.E.2d 656 (1991); Tappe v. Iowa Methodist Medical Center, 477 N.W.2d 396 (Ia.1991); Rollins v. Peterson, 813 P.2d 1156 (Utah 1991); Walton v. Potlatch Corp., 116 Idaho 892, 781 P.2d 229 (1989); Toy v. District of Columbia, 549 A.2d 1 (D.C.App.1988); Wickwire v. Arctic Circle Air Services, 722 P.2d 930 (Alaska 1986); Streich v. Hilton-Davis, 214 Mont. 44, 692 P.2d 440, 449 (1984); Freund v. Cellofilm Properties, Inc., 87 N.J. 229, 432 A.2d 925 (1981); Marlow v. City of Columbia Heights, 284 N.W.2d 389 (Minn.1979); Petition of Kinsman Transit Co., 338 F.2d 708, 721-26 (2d Cir.1964). “The common law is not a brooding omnipresence in the sky, but the articulate voice of some sovereign or quasi sovereign that can be identified.” Southern Pacific Co. v. Jensen, 244 U.S. 205, 222, 37 S.Ct. 524, 531, 61 L.Ed. 1086 (1917) (Holmes, J., dissenting). The voices of the quasi-sovereigns that are the states of the United States sing negligence "with a different pitch.
The “serendipity” theory advanced by the plaintiffs in Wadleigh is that if the defendants did not do enough to protect hemophiliacs from the risk of Hepatitis B, they are liable to hemophiliacs for any consequences — including infection by the more dangerous and at the time completely unknown AIDS virus — that proper measures against Hepatitis B would, all unexpectedly, have averted. This theory of liability, which draws support from Judge Friendly’s opinion in Petition of Kinsman Transit Co., supra, 338 F.2d at 725, dispenses, rightly or wrong- ■ ly from the standpoint of the Platonic Form of negligence, with proof of foreseeability, even though a number of states, in formulating their tests for negligence, incorporate the foreseeability of the risk into the test. See, e.g., Fawley v. Martin’s Supermarkets, Inc., 618 N.E.2d 10, 13 (Ind.App.1993); Comment Note, “Foreseeability as an Element of Negligence and Proximate Cause,” 100 A.L.R.2d 942 (1994). These states follow Judge Cardozo’s famous opinion in Palsgraf v. Long Island R.R., 248 N.Y. 339, 162 N.E. 99 (1928), under which the HIV plaintiffs might (we do not say would — we express no view on the substantive issues in this litigation) be barred from recovery on the ground that they were unforeseeable victims of the alleged failure of the defendants to take adequate precautions against infecting hemophiliacs with Hepatitis B and that therefore the drug companies had not violated any duty of care to them.
The plaintiffs’ second theory focuses on the questions when the defendants should have learned about the danger of HIV in the blood supply and when, having learned about it, they should have taken steps to eliminate the danger or at least warn hemophiliacs or their physicians of it. These questions also may be sensitive to the precise way in which a state formulates its standard of negligence. If not, one begins to wonder why this country bothers with different state legal systems.
Both theories, incidentally, may be affected by differing state views on the role of industry practice or custom in determining the existence of negligence. In some states, the standard of care for a physician, hospital, or other provider of medical services, including blood banks, is a professional standard, that is, the standard fixed by the relevant profession. In others, it is the standard of ordinary care, which may, depending on judge or jury, exceed the professional standard. Joseph Kelly, “The Liability of Blood Banks and Manufacturers of Clotting Products to Recipients of HIV-Infected Blood: A Comparison of the Law and Reaction in the United States, Canada, Great Britain, Ireland, and Australia,” 27 John Marshall Law Review 465, 472-74 (1994); United Blood Services v. Quintana, 827 P.2d 509, 525-26 (Colo.1992). Which approach a state follows, and whether in those states that follow the professional-standard approach manufacturers of blood solids would be assimilated to blood banks as providers of medical services [1302]*1302entitled to shelter under the professional standard, could make a big difference in the liability of these manufacturers. We note that persons infected by HIV through blood transfusions appear to have had little better luck suing blood banks than HIV-positive hemophiliacs have had suing the manufacturers of blood, solids. Kelly, supra, at 465-77.
The diversity jurisdiction of the federal courts is, after Erie, designed merely to provide an alternative forum for the litigation of state-law claims, not an alternative system of substantive law for diversity cases. But under the district judge’s plan the thousands of members of the plaintiff class will have their rights determined, and the four defendant manufacturers will have their duties determined, under a law that is merely an amalgam, an averaging, of the nonidentical negligence laws of 51 jurisdictions. No one doubts that Congress could constitutionally prescribe a uniform standard of liability for manufacturers of blood solids. It might we suppose promulgate pertinent provisions of the Restatement (Second) of Torts. The point of Erie is that Article III of the Constitution does not empower the federal courts to create such a regime for diversity cases.
If in the course of individual litiga-tions by HIV-positive hemophiliacs juries render special verdicts that contain findings which do not depend on the differing state standards of negligence—for example a finding concerning the date at which one or more of the defendants learned of the danger of HIV contamination of the blood supply— these findings may be given collateral estop-pel effect in other lawsuits, at least in states that allow “offensive” use of collateral estop-pel. In that way the essential purpose of the class action crafted by Judge Grady will be accomplished. If there are relevant differences in state law, findings in one suit will not be given collateral estoppel effect in others, Commissioner v. Sunnen, 333 U.S. 591, 599-600, 68 S.Ct. 715, 720-21, 92 L.Ed. 898 (1948); Goodson v. McDonough Power Equipment, Inc., 2 Ohio St.3d 193, 2 OBR 732, 443 N.E.2d 978, 987-88 (1983)—and that is as it should be.
The plaintiffs argue that an equally important purpose of the class certification is to overcome the shyness or shame that many people feel at acknowledging that they have AIDS or are HIV-positive even when the source of infection is not.a stigmatized act. That, the plaintiffs tell us, is why so few HIV-positive hemophiliacs have sued. We do not see how a class action limited to a hand-fill of supposedly common issues can alleviate that problem. Any class member who wants a share in any judgment for damages or in any settlement will have to step forward at some point and identify himself as having AIDS or being HIV-positive. He will have to offer jury findings as collateral estoppel, overcome the defendants’ defenses to liability (including possible efforts to show that the class member became infected with HIV through a source other than the defendants’ product), and establish his damages. If the privacy of these class members in these follow-on proceedings to the class action is sought to be protected by denominating them “John Does,” that is something that can equally well be done in individual lawsuits. The “John Doe” device—and with it the issue of privacy—is independent of class certification.
The third respect in which we believe that the district judge has exceeded his authority concerns the point at which his plan of action proposes to divide the trial of the issues that he has certified for class-action treatment from the other issues involved in the thousands of actual and potential claims of the representatives and members of the class. Bifurcation and even finer divisions of lawsuits into separate trials are authorized in federal district courts. Fed.R.Civ.P. 42(b); Sellers v. Baisier, 792 F.2d 690, 694 (7th Cir.1986). And a decision to employ the procedure is reviewed deferentially. Berry v. Deloney, 28 F.3d 604, 610 (7th Cir.1994); De Witt, Porter, Huggett, Schumacher & Morgan, S.C. v. Kovalic, 991 F.2d 1243, 1245 (7th Cir.1993); Barr Laboratories, Inc. v. Abbott Laboratories, 978 F.2d 98, 105 (3d Cir.1992). However, as we have been at pains to stress recently, the district judge, must carve at the joint. Hydrite Chemicals Co. v. Calumet Lubricants Co., 47 F.3d 887, 890-91 (7th Cir.1995); cf. McLaughlin v. State Farm Mutual Automobile Ins. Co., 30 [1303]*1303F.3d 861, 870-71 (7th Cir.1994). Of particular relevance here, the judge must not divide issues between separate trials in such a way that the same issue is reexamined by different juries. The problem is not inherent in bifurcation. It does not arise when the same jury is to try the successive phases of the litigation. But most of the separate “cases” that compose this class action will be tried, after the initial trial in the Northern District of Illinois, in different courts, scattered throughout the country. The right to a jury trial in federal civil cases, conferred by the Seventh Amendment, is a right to have juria-ble issues determined by the first jury impaneled to hear them (provided there are no errors warranting a new trial), and not reexamined by another finder of fact. This would be obvious if the second finder of fact were a judge. Byrd v. Blue Ridge Rural Electric Cooperative, Inc., 356 U.S. 525, 537-38, 78 S.Ct. 893, 900-01, 2 L.Ed.2d 953 (1958); Davenport v. DeRobertis, 844 F.2d 1310, 1313-14 (7th Cir.1988); Hunter v. Allis-Chalmers Corp., 797 F.2d 1417, 1421 (7th Cir.1986). But it is equally true if it is another jury. Gasoline Products Co. v. Champlin Refining Co., 283 U.S. 494, 500, 51 S.Ct. 513, 515, 75 L.Ed. 1188 (1931); McDaniel v. Anheuser-Busch, Inc., 987 F.2d 298, 305 (5th Cir.1993); Alabama v. Blue Bird Body Co., 573 F.2d 309, 318 (5th Cir.1978). In this limited sense, a jury verdict can have collateral estoppel effect. Davenport v. DeR-obertis, supra, 844 F.2d at 1313-14; Hudson Ins. Co. v. City of Chicago Heights, 48 F.3d 234, 237-38 (7th Cir.1995).
The plan of the district judge in this case is inconsistent with the principle that the findings of one jury are not to be reexamined by a second, or third, or nth jury. The first jury will not determine liability. It will determine merely whether one or more of the defendants was negligent under one of the two theories. The first jury may go on to decide the additional issues with regard to the named plaintiffs. But it will not decide them with regard to the other class members. Unless the defendants settle, a second (and third, and fourth, and hundredth, and conceivably thousandth) jury will have to decide, in individual follow-on litigation by class members not named as plaintiffs in the Wad-leigh case, such issues as comparative negligence — did any class members knowingly continue to use unsafe blood solids after they learned or should have learned of the risk of contamination with HIV? — and proximate causation. Both issues overlap the issue of the defendants’ negligence. Comparative negligence entails, as the name implies, a comparison of the degree of negligence of plaintiff and defendant. See, e.g., Alaska Stat. § 09.17.080; ILCS 735 5/2-1116; N.J.Stat. § 2A:15-5.1; Ohio Rev.Code' § 2315.19; Utah Code § 78-27-38. Proximate causation' is found by determining whether the harm to the plaintiff followed in some sense naturally, uninterruptedly, and with reasonable probability from the negligent act of the defendant. It overlaps the issue of the defendants’ negligence even when the state’s law does not (as many states do) make the foreseeability of the risk to which the defendant subjected the plaintiff an explicit ingredient of negligence. See, e.g., Powell v. Drumheller, — Pa.-, 653 A.2d 619 (1995); Whittaker v. Saraceno, 418 Mass. 196, 635 N.E.2d 1185 (1994); Vincent v. Fairbanks Memorial Hospital, 862 P.2d 847, 851-52 (Alaska 1993); Flight Line, Inc. v. Tanksley, supra, 608 So.2d at 1158-59; Wasfi v. Chaddha, 218 Conn. 200, 588 A.2d 204 (1991). A second or subsequent jury might find that the defendants’ failure to take precautions against infection with Hepatitis B could not be thought the proximate cause of the plaintiffs’ infection with HIV, a different and unknown blood-borne virus. How the resulting inconsistency between juries could be prevented escapes us.
The protection of the right conferred by the Seventh Amendment to trial by jury in federal civil cases is a traditional office of the writ of mandamus. Beacon Theatres v. Westover, 359 U.S. 500, 510-11, 79 S.Ct. 948, 956-57, 3 L.Ed.2d 988 (1959); Dairy Queen, Inc. v. Wood, 369 U.S. 469, 472, 82 S.Ct. 894, 897, 8 L.Ed.2d 44 (1962); Maloney v. Plunkett, supra; First National Bank v. Warren, 796 F.2d 999 (7th Cir.1986). When the writ is used for that purpose, strict compliance with the stringent conditions on the availability of the writ (including the requirement of proving irreparable harm) is excused. In [1304]*1304Beacon, for example, if the judge had gone ahead and tried the equitable claims first and made his decision collateral estoppel in the subsequent jury trial, the losing party would have been entitled to a new trial, wiping out the judge’s decision. There was no irreparable harm, yet mandamus was granted— which is one reason why in eases like Malo-ney and Sandahl we have said that the use of the writ cannot be reduced to formula. But the looming infringement of Seventh Amendment rights is only one of our grounds for believing this to be a ease in which the issuance of a writ of mandamus is warranted. The others as we have said are the undue and unnecessary risk of a monumental industry-busting error in entrusting the determination of potential multi-billion dollar liabilities to a single jury when the results of the previous cases indicate that the defendants’ liability is doubtful at best and the questionable constitutionality of trying a diversity case under a legal standard in force in no state. We need not consider whether any of these grounds standing by itself would warrant mandamus in this case. Together they make a compelling case.
We know that an approach similar to that proposed by Judge Grady has been approved for asbestos litigation. See in particular Jenkins v. Raymark Industries, Inc., 782 F.2d 468 (5th Cir.1986); In re School Asbestos Litigation, 789 F.2d 996 (3d Cir.1986). Most federal courts, however, refuse to permit the use of the class-action device in mass-tort cases, even asbestos cases. Thomas E. Willging, Trends in Asbestos Litigation 93-98 (Federal Judicial Center 1987); cf. In re Fibreboard Corp., supra; In re Joint Eastern & Southern District Asbestos Litigation, 982 F.2d 721 (2d Cir.1992). Those courts that have permitted it have been criticized, and alternatives have been suggested which recognize that a sample of trials makes more sense than entrusting the fate of an industry to a single jury. See, e.g., Michael J. Saks & Peter David Blanck, “Justice Improved: The Unrecognized Benefits of Aggregation and Sampling in the Trial of Mass Torts,” 44 Stan.L.Rev. 815 (1992). The number of asbestos cases was so great as to exert a well-nigh irresistible pressure to bend the normal rules. No comparable pressure is exerted by the. HIV-hemophilia litigation. That litigation can be handled in the normal way without undue inconvenience to the parties or to the state or federal courts.
The defendants have pointed out other serious problems -with the district judge’s plan, but it is unnecessary to discuss them. The petition for a writ of mandamus is granted, and the district judge is directed to decertify the plaintiff class.