Judge JACOBS concurs in the opinion of the Court, and also files a separate opinion.
CALABRESI, Circuit Judge:
Appellants, a group of more than fifty insurers, appeal from a declaratory judgment of the United States District Court for the Southern District of New York (John S. Mar[928]*928tin, Jr., Judge) following a jury trial on the insurance policies provided by Appellants to Appellee E.R. Squibb & Sons, Inc. (“Squibb”)- Although this case has been litigated in federal court for sixteen years, the existence of diversity jurisdiction pursuant to 28 U.S.C. § 1332 (1994) has not been established. This is due to the presence in the suit of “Certain Underwriters at Lloyd’s of London” (“Lloyd’s”).1 Accordingly, we vacate the judgment and remand to the district court for further proceedings on the jurisdictional issue.
BACKGROUND
In 1982, Squibb brought a declaratory action against its domestic primary and excess insurers seeking indemnification for product liability claims arising out of the use of the drug diethylstilbestrol (“DES”). In 1984, Squibb filed a new consolidated complaint against its insurers that included a host of new domestic and foreign defendants.2 Among those named in the consolidated complaint was defendant Allen Peter Dennis Haycock (“Haycock”), a British subject, who was named “as a representative underwriter representing certain underwriters at Lloyd’s [ of] London, being all underwriters who subscribed the policies of insurance issued to the plaintiff and upon which the plaintiff brings the present action.” The parties subsequently stipulated that Haycock was “appearing in this action in his individual capacity, and for administrative convenience, as a representative of all Lloyd’s Underwriters.” Squibb, which is a citizen of Delaware and New Jersey, invoked diversity jurisdiction and New York law to support its claims. For the next thirteen years, the ease proceeded with discovery and then culminated in a jury trial in which Squibb prevailed.3
While the appeal of Squibb’s judgment was pending, this court decided Advani Enterprises, Inc. v. Underwriters at Lloyds, 140 F.3d 157 (2d Cir.1998). Advani raised, but did not decide, the difficult questions involved in determining how the Lloyd’s underwriters should be analyzed for purposes of the requirements of diversity jurisdiction.4 See id. at 160-61. In light of the discussion in Advani, this court raised sua sporite the issue of whether, due to Haycock’s status as a defendant representing “Certain Underwriters at Lloyd’s of London,” diversity jurisdiction existed in the district court. We requested additional briefing on that question.5
In the course of that briefing, counsel discovered that Haycock had recently died. Squibb and Lloyd’s now move to add Stephen Merrett, another Lloyd’s underwriter and British subject, perhaps in replacement of Haycock,6 and to find that diversity jurisdiction exists in this action.
Before addressing the complex jurisdictional issues raised by this case, we think it helpful to summarize our understanding of the unique structure that is Lloyd’s of Lon-
[929]*929don. Lloyd’s began as a coffee house but has developed into one of the world’s leading markets for insurance. See Godfrey Hodgson, Lloyd’s of London 49-75 (1984). This market, however, operates in accordance with age-old customs that are, to say the least, unusual in American business law.
The anonymous underwriters of Lloyd’s insurance, who are commonly referred to as “Names,” invest in a percentage of the policy risk. While the rewards of a Lloyd’s investment can be great, each Lloyd’s Name is exposed to unlimited liability, but only for his or her share of the loss on a policy that the Name has underwritten. In other words, the liability of each Name on any given policy, while unlimited, is several and not joint. Insurance from Lloyd’s is typically subscribed to by hundreds of Names belonging to different subgroups known as “syndicates.” Although syndicates within the Lloyd’s market negotiate with each other to spread insurance risk, the syndicates themselves have been said to have no independent legal identity. See The Society of Lloyd’s v. Clementson, [1996] 5 Re. L.R. 215.
The syndicate Names do not manage their own investments. Instead, each syndicate appoints one of its Names (who is usually an insurance broker) to represent the collective interests of the Names in that syndicate. This person is known as the “lead underwriter.” The typical Lloyd’s policy contains a clause providing that “any [Name] can appear as representative of all [Names].” In practice, however, since many Names from various syndicates are usually involved in any particular policy, the lead underwriter from one of the underwriting syndicates is designated as the representative of all the Names in any of the relevant syndicates, and he is the only Name disclosed on a policy.
When litigation over a Lloyd’s policy occurs, only one Name (the lead underwriter disclosed on the policy) is ordinarily sued. Nevertheless, all the Names subscribing to that policy are liable for their several shares of any adverse judgment against the Lloyd’s underwriters. This is because the standard Lloyd’s policy running between the insured and each Name states “that in any suit instituted against any one of [the Names] upon this contract, [all the Names] will abide by the final decision of such Court or of any Appellate Court in the event of an appeal.” Each Name is, therefore, bound by contract with the insured to adhere to the decision reached in the suit.7
DISCUSSION
I
Since “subject matter jurisdiction is an unwaivable sine qua non for the exercise of federal judicial power,” Curley v. Brignoli, Curley & Roberts Assocs., 915 F.2d 81, 83 (2d Cir.1990), we must consider whether there is diversity jurisdiction in this action, even though the parties have assumed for the past sixteen years that they had a right to be in federal court, and desperately — vehemently even — want the court to find that jurisdiction exists. We understand and fully sympathize with that desire. Sixteen years is a long time. And there is no suggestion that the parties engaged in any collusion or misbehavior to create federal jurisdiction when this suit was originally brought. It would, therefore, be most unfortunate if, in the end, it were necessary for us to hold that no federal jurisdiction existed and that the parties had to start all over again in state court.
Having said that, we must nevertheless recognize that jurisdiction is not a game. As the Supreme Court has made abundantly clear, it is one of the fundamental tenets of our Constitution that only some eases may be brought in federal court. See Healy v. Ratio, 292 U.S. 263, 270, 54 S.Ct. 700, 78 L.Ed. 1248 (1934) (“Due regard for the rightful independence of state governments, which should actuate federal courts, requires that they scrupulously confine their own jurisdiction to the precise limits which the [diversity] statute has defined.”). And no amount of agreement by the parties can create jurisdiction where none exists. See Philippines v. Marcos, 806 F.2d 344, 352 (2d Cir.1986).
Nor should we, to accommodate the particular and unfortunate situation before us, make law on jurisdiction that would be undesirable in the mass of eases. We must, [930]*930moreover, carefully follow the guidelines on jurisdiction that the Supreme Court has set forth. Nobody’s interest would be served if we by stretching the law found jurisdiction to exist, only to have that position ultimately rejected by the High Court. If it would be dismal for all concerned to have to go back to square one after sixteen years, it would be even more unfortunate were we to create a situation that, while seemingly determining the matter, would instead result in the parties being forced to return to the beginning after eighteen or twenty years.
With that as a prologue, we turn to the jurisdictional problems this case presents. It is axiomatic that, for diversity jurisdiction to be available, all of the adverse parties in a suit must be completely diverse with regard to citizenship. See Strawbridge v. Curtiss, 7 U.S. (3 Cranch) 267, 2 L.Ed. 435 (1806). In the ease before us, we have been assured that such complete diversity of citizenship is clearly present as to every party except the Lloyd’s underwriters. Haycock— who, though dead, still remains in this case—• was sued by Squibb both as an individual Name and as the lead underwriter representing all of the Lloyd’s Names on Squibb’s policy.8 The first question we must therefore decide is whether only the Lloyd’s representative lead underwriter needs to meet the diversity requirement or whether each and every Name represented (i.e. each Name subscribing to the Lloyd’s policy at issue in the case) must be of diverse citizenship.
If the lead underwriter is the proper focal point, then the parties assert that there is no jurisdictional problem in the case. They allege—and there is no evidence in the record to the contrary—that Haycock is a foreign subject who could be sued without destroying complete diversity. See 28 U.S.C. § 1332(a)(3) (1994) (providing that diversity jurisdiction exists in civil suits between “citizens of different States and in which citizens or subjects of a foreign state are additional parties”). On the other hand, if the citizenship of each Name must be considered, then a remand would clearly be required in this suit as currently constituted, since the record does not disclose the identity, let alone the citizenship, of the Names involved in the case. And without knowledge of that citizenship, it would be impossible to say that complete diversity exists.
The test for determining the existence of diversity jurisdiction is generally “the citizenship of real parties to the controversy.” See Navarro Savings Ass’n v. Lee, 446 U.S. 458, 461, 100 S.Ct. 1779, 64 L.Ed.2d 425 (1980). Since the Lloyd’s market presents a unique structure for jurisdictional analysis, it is not surprising that the two circuits that have addressed issues analogous to the one that we now face have reached opposite conclusions.9 Compare Certain Interested Underwriters at Lloyd’s, London v. Layne, 26 F.3d 39 (6th Cir.1994) (holding that a lead underwriter is the only real party of interest to the controversy and hence is the only Lloyd’s Name that must be completely diverse),10 with Indiana Gas Co. v. Home Ins. Co., 141 F.3d 314 (7th Cir.) (holding that each and every Name must meet the complete diversity rule), cert. denied — U.S. -, 119 S.Ct. 339, 142 L.Ed.2d 280 (1998).11
[931]*931As the suit is currently constituted, we believe that the Seventh Circuit has the better of this argument. We therefore reject the notion that Squibb’s suit, in its present posture, allows our jurisdictional inquiry to focus solely on the citizenship of a Lloyd’s representative lead underwriter. The general rule undoubtedly is “that federal courts must look to the individuals being represented rather than their collective representative to determine whether diversity of citizenship exists.” See Northern Trust Co. v. Bunge Corp., 899 F.2d 591, 594 (7th Cir.1990) (finding that a representative of a group of shareholders took on the citizenship, for jurisdictional purposes, of each shareholder represented). The question before us, therefore, is: can the Lloyd’s underwriters fit into one of the exceptions to that rule?
II
There are three principal exceptions. The first involves corporations, whose citizenship alone matters and the status of whose individual shareholders can therefore be ignored. See, e.g., Airlines Reporting Corp. v. S & N Travel, Inc., 58 F.3d 857, 861 (2d Cir.1995). That exception is not relevant to this case. The second possibility concerns trusts, as to which the Supreme Court has deemed the citizenship of the trustees to be determinative. See Navarro, 446 U.S. at 465-66, 100 S.Ct. 1779. The contention— strongly urged by the parties—that the rule of Navarro should apply in the instant ease because the Lloyd’s market is a trust and the lead underwriter is a trustee is not persuasive.
As the Seventh Circuit stated in Indiana Gas:
Trustees own the corpus; ownership is what distinguishes a trustee from an agent. [Lead] underwriters commit the wealth of the syndicates’ names to fulfilment of the policy, which makes the insurance valuable to the customer, but the [lead] underwriters do not own this wealth or exercise over it any dominion other than the power to underwrite risks.
Indiana Gas, 141. F.3d at 318. We agree with this analysis and conclude that a Lloyd’s lead underwriter cannot be deemed a trustee for jurisdictional purposes.
The final exception involves class actions, where the Court has held that only the citizenship of the class representatives is relevant for the establishment of diversity jurisdiction. See Snyder v. Harris, 394 U.S. 332, 340, 89 S.Ct. 1053, 22 L.Ed.2d 319 (1969) (citing Supreme Tribe of Ben-Hur v. Cauble, 255 U.S. 356, 41 S.Ct. 338, 65 L.Ed. 673 (1921)) And—despite the fact that this case was not originally brought as a class action— the parties understandably seek to invoke this exception. They ask the court to rechar-acterize Squibb’s complaint as stating a Rule 23.2 class action12 against the Lloyd’s underwriters pursuant to our power to correct defective allegations of jurisdiction, see 28 U.S.C. § 1653 (1994).13 The parties contend that if we read Squibb’s complaint as a Rule 23.2 class action, only Haycock’s citizenship will be relevant and diversity jurisdiction can be found. See Snyder, 394 U.S. at 340, 89 S.Ct. 1053.
They assert, moreover, that we have done just this in a prior case. In Curley, they say, we exercised our discretion and recharacter-ized a derivative suit by various limited partners against their partnership as a Rule 23.2 class action in order to salvage diversity jurisdiction. See Curley, 915 F.2d at 82, 87-88. The plaintiffs in Curley were the victims of a breach of fiduciary trust by the general partner. See id. at 83, This breach had included a significant misappropriation of the partnership’s assets. See id. While judgment for the plaintiffs was being appealed, the Supreme Court decided Carden v. Arkoma Associates., 494 U.S. 185, 110 S.Ct. 1015, 108 L.Ed.2d 157 (1990). See Curley, 915 F.2d at 82. Carden held that each and every partner in a limited partnership had to be diverse in a suit by or against a partnership. See Carden, 494 U.S. at 195-96, 110 S.Ct. 1015. The defrauding general partner in Curley hastily pointed out that complete diversity did not [932]*932exist between the plaintiffs and each and every limited partner. See id. at 83-84. We concluded that, since “[d]efendants [were] not prejudiced by the proposed amendment,” the plaintiffs’ action could be recast as a Rule 23.2 class action to sustain diversity jurisdiction. See id. at 88.
The parties’ request in this case presents us with a somewhat different situation. First, unlike Curley, this suit was before the district court after Cardenwas decided, and the parties, therefore, had an opportunity to amend their complaint so as to state a class action claim if they wished to do so. Second, some of the equitable considerations present in Curleyiwhere we were trying to prevent the perpetuation of a fraud against the limited partners) are absent in the case before us.
More important, when faced with a similar question of whether a suit against the Lloyd’s underwriters could be reeonstrued as a Rule 23.2 class action, the Seventh Circuit, in Indiana Gas, rejected the idea on its merits. See Indiana Gas, 141 F.3d at 321 (petition for rehearing). Because it considered the Lloyd’s syndicates to be the functional equivalent of limited partnerships, the Seventh Circuit concluded that converting a suit against the Lloyd’s underwriters into a class action would undermine the substantive distinction, for purposes of jurisdiction, between Rule 23 classes and limited partnerships.14 See id.
Admittedly, this distinction is not pellucid. And, at a functional level, the decision in Carden, 494 U.S. at 195-96, 110 S.Ct. 1015, requiring courts to look at the citizenship of all the members of a limited partnership in order to establish diversity jurisdiction cannot be easily reconciled with Snyder, 394 U.S. at 340, 89 S.Ct. 1053, under which a Rule 23.2 class representative is the sole focus of the jurisdictional inquiry. Thus, as Justice O’Connor (arguing that Carden was an exercise in formalism) noted in her dissent: “[DJespite the Court’s holding ..., unincorporated associations may gain access to the federal courts by bringing or defending suit as a Rule 23 class action, in which case the citizenship of the members of the class would not be considered.” See Carden, 494 U.S. at 207, 110 S.Ct. 1015 (O’Connor, J., dissenting).
Justice O’Connor did not, of course, suggest that recharacterizing such eases retroactively was appropriate, and her dissent must be taken to speak only to cases initially brought as class actions. Nevertheless, her assertion was left unanswered by the Carden majority, and so the apparent tension between class actions and limited partnerships in the Supreme Court’s diversity jurisprudence remains unresolved.
The Seventh Circuit addressed this tension in Indiana Gas and rejected Justice O’Con-nor’s suggestion. Indiana Gas held that the plaintiffs could not evade the rule of Carden by reclassifying the suits against the Lloyd’s syndicates as class actions since “a rule of procedure [cannot] expand the subject-matter jurisdiction of the federal courts.” See Indiana Gas, 141 F.3d at 321 (petition for rehearing); see also Fed.R.Civ.P. Rule 82 (“These rules shall not be construed to extend or limit the jurisdiction of the United States district courts or the venues of actions therein.”).
There are potentially significant differences between Indiana Gas and the instant case. The suit in Indiana Gas, unlike the action in this ease, was brought against a Lloyd’s lead underwriter only in his representative capacity. See Indiana Gas, 141 F.3d at 321 (petition for rehearing). And the Seventh Circuit considered the analogy to class actions inapt because the court asserted that the proposed Lloyd’s Rule 23.2 representative — the lead underwriter — did not have a personal stake in the outcome as an ordinary class representative does and must. See id. By contrast, Squibb has made a direct claim against Haycock, whom it has sued both as an individual Name and as a representative of all the Names. Actually, given the fact that under the Lloyd’s contracts all Names on a policy agree to liability whenever the Name sued has been adjudged liable, we believe that even if Haycock had [933]*933been sued only as a representative he would—as a Name—have possessed the individual stake necessary to act as an appropriate Rule 23.2 class representative.
Since the Seventh Circuit based its decision in Indiana Gas on a contrary assumption—that the lead underwriter had no such individual stake—its decision not to recharac-terize need not be read as denying the possibility of recharacterization where such a stake exists. We cannot deny, however, that the affect, if not the technical holding, of Indiana Gas suggests that the Seventh Circuit looks unfavorably on recharacteriza-tions.15 Under the circumstances, we should be hesitant to consider whether this case is subject to recharacterization unless the district court rules that it is otherwise without jurisdiction, and also that this ease both can and should be recharacterized as a class action.
There is, moreover, an additional reason why a simple recharacterization of this case as a class action is not at this time advisable. Recasting the case as a class action with Haycock as the class representative would not, without much more, solve the jurisdictional problems we face. It would only force us to confront difficult questions involving the dollar amount in controversy in this suit. The requirement that a certain amount be at stake is an unwaivable jurisdictional one, see 28 U.S.C. § 1332(a).16 And whether in the instant case the jurisdictional amount is met depends on how we interpret a series of complex and ambiguous rules.
Thus, even if it were established that Haycock as class representative met the jurisdictional amount as to his own individual liability (which is itself anything but obvious),17 that would not end the matter. For—except as possibly modified by 28 U.S.C. § 1367 and 28 U.S.C. § 1332(a)(3), which we discuss below—each and every Name would still have to satisfy the jurisdictional amount.18
It is well established that for diversity actions “the rule applicable to several plaintiffs having separate claims, that each must represent an amount sufficient to give the court jurisdiction, is equally applicable to several liabilities of different defendants to the same plaintiff.” See Walter v. Northeastern R.R. Co., 147 U.S. 370, 374, 13 S.Ct. 348, 37 L.Ed. 206 (1893) (emphasis added). And the fact that the several liability of the Names derives from a single insurance policy does not alter the jurisdictional analysis. See Motorists Mut. Ins. Co. v. Simpson, 404 F.2d 511, 513 (7th Cir.1968) (“When two or more claimants are joined as defendants in one declaratory judgment suit, the pecuniary test of jurisdiction depends on whether the potential liability of the insurer is joint or several.... That the foundation of both potential claims is based on provisions in the same document is immaterial.”).
Nor is a plaintiff permitted to aggregate its claims against individual defendants to meet the jurisdictional requirement. Aggregation to achieve diversity jurisdiction is barred when the liability of the defendants is several and not joint. See Walter, 147 U.S. at 373, 13 S.Ct. 348; see also Thomson v. Gaskill, 315 U.S. 442, 447, 62 S.Ct. 673, 86 L.Ed. 951 (1942) (“Aggregation of plaintiffs’ claims cannot be made merely because the claims are derived from a single instrument....”). As a result, each and every severally liable defendant must, in the normal course of things, meet the amount in controversy.
[934]*934It follows that viewing Squibb’s suit as a Rule 23.2 class action would not seem readily to solve the problem of jurisdictional amount. But the parties argue that it does. They contend, for example, that § 1332(a)(3) exempts foreign nationals from the jurisdictional minimum. That is a question that no court has ever addressed, and reaching the conclusion urged by the parties is not without significant problems. See generally Note, Federal Jurisdiction over Suits Between Diverse United States Citizens with Aliens Joined to Both Sides of the Controversy Under 28 U.S.C. § 1332(a)(3), 38 Rutgers L.Rev. 71, 87-97 (1985) (describing the legislative history of § 1332(a)(3) as focused exclusively on the citizenship diversity question and not on jurisdictional amount).
Moreover, accepting the parties’ contention would still not mean that the jurisdictional problems we face would disappear. For even if § 1332(a)(3) means what the parties assert that it does, the provision exempts only foreign nationals. Thus, while Haycock (the purported original class representative and a British subject) would, on such a reading, undoubtedly satisfy the jurisdictional requirement, that would not mean that all the individual Names met the amount. Since we do not know the identity, let alone the nationality, of these Names, we would still be left without the requisite certainty that the jurisdictional minimum was satisfied by those to whom it manifestly applies — U.S. nationals. It follows that, unless class action members who are severally liable do not need to meet the jurisdictional amount individually for some other reason, then reading § 1332(a)(3) as the parties ask us to do would not suffice to establish jurisdiction.
We must therefore turn to the parties’ alternative arguments with respect to the jurisdictional amount. In Zahn v. International Paper Co., 414 U.S. 291, 94 S.Ct. 505, 38 L.Ed.2d 511 (1973), the Supreme Court expressly barred aggregation of several claims in class actions and held that each defendant, even if seen as part of a class, would have to meet the jurisdictional amount requirement individually. But the parties contend (1) that § 1367 overturns Zahn and now permits aggregation in all class action suits;19 and (2) that Squibb’s claims against the Names can be aggregated in a Rule 23.2 action even if Zahn still governs and aggregation is barred in other types of class actions.
Two circuits have indeed held that, after the enactment of § 1367, Zahn is no longer good law. See Stromberg Metal Works, Inc. v. Press Mechanical, Inc., 77 F.3d 928, 931 (7th Cir.1996); Free v. Abbott Lab. (In re Abbott Lab.), 51 F.3d 524, 529 (5th Cir.1995). This conclusion was based on the fact that § 1367 does not exempt Rule 23 actions from its reach, while specific exceptions are made for other portions of the Federal Rules of Civil Procedure. See Abbott Lab., 51 F.3d at 527. No court has held, however, that such an overturning of Zahn is applicable to suits (such as this one) brought before 1990 when § 1367 went into effect. Cf. More v. Intelcom Support Servs., Inc., 960 F.2d 466, 473 (5th Cir.1992) (noting that § 1367 “affects only cases filed on or after December 1, 1990”). Moreover, a lower court in our circuit has asserted that the notion that § 1367 has overturned Zahn at all “is the minority view and has not been followed in this Circuit.” See Bernard v. Gerber Food Prods. Co., 938 F.Supp. 218, 223 (S.D.N.Y.1996). In fact, this court has not squarely addressed the issue, and the Supreme Court has never considered it. Under the circumstances, we believe that a decision by us on the question at this time would be unwise.
. The parties’ argument that aggregation is permitted in Rule 23.2 class actions, notwithstanding Zahn, is also a matter of first impression, not only in this circuit but in every circuit. There is some authority suggesting [935]*935that “[s]inee [a Rule 23.2] suit was recognized as a true class suit under former Rule 23(a), it should follow that the common claims by or against the individual members of the unincorporated association can be aggregated to sustain jurisdictional amount requirements.” See 3B James Wm. Moore et al., Moore’s Federal Practice ¶ 23.2.02 at 23.2-7 to 23.2-8 (2d ed.1985) (footnote omitted). But no court has ever held that aggregation is permitted in a Rule 23.2 action.20 And, once again, we deem it'inadvisable to decide this novel question at present.
To summarize: Recasting this suit as a class action and finding that, as such, jurisdiction exists presents substantial, though perhaps not insuperable, difficulties.21 It runs against the affect, if not the strict holding, of the Seventh Circuit, which seems to believe that recharacterization opens a detour around Carden. And it would only resolve the jurisdictional problem in this case if we made a series of other holdings on issues of first impression in this circuit as to which we have no Supreme Court guidance. For all these reasons, we believe that deciding the validity of the class action approach and considering the subsidiary questions it entails, .should, in this case, be undertaken only as a last resort, either by us or by the district court. Accordingly—though mindful of our action in Curley—we decline at this time to reconstrue the instant case as a class action. In doing so, we express no opinion as to whether it could ultimately be so recast, or on the subsidiary issues that would then need to be decided. Nor do we foreclose the district court—if all other avenues of finding jurisdiction fail—from addressing these questions in the first instance.22
Ill
We therefore turn to the parties’ alternative suggestion that this court save diversity jurisdiction by adding Merrett as an individual Name defendant and dismissing Haycock (at least, in his representative capacity) and retaining him in this suit (if at all) only as an individual Name. In this respect, the parties rely on Newman-Green, Inc. v. Alfonzo-Larrain, 490 U.S. 826, 109 S.Ct. 2218, 104 L.Ed.2d 893 (1989), where the Supreme Court held that, even on appeal, a court has the power to dismiss a dispensable nondi-verse party from a suit under Rule 21 of the Federal Rules of Civil Procedure in order to salvage jurisdiction. See id. at 832-33, 109 S.Ct. 2218 (extending Rule 21 of the Federal Rules of Civil Procedure to appellate courts). The Supreme Court has recently stated that “Once a diversity suit has been tried in federal court ... considerations of finality, efficiency, and economy become overwhelming.” See Caterpillar Inc. v. Lewis, 519 U.S. 61, -, 117 S.Ct. 467, 476, 136 L.Ed.2d 437 (1996). And it has emphasized that “requiring dismissal after years of litigation would impose unnecessary and wasteful burdens on the parties, judges, and other litigants waiting for judicial attention.” Newman-Green, 490 U.S. at 836, 109 S.Ct. 2218. We approach the parties’ suggestions with respect to Haycock and Merrett with these admonitions in mind.
As a preliminary matter, we must address the question of capacity because so much may ride on the distinction between a suit against Haycock or Merrett in then-representative capacities and one against them as individuals. Capacity to' sue and be sued in federal court is governed by Fed. R.Civ.P. 17(b). The relevant portion of that [936]*936Rule says that a person’s capacity to be sued as an individual (as opposed to as a representative) depends on the law of the person’s domicile. Lack of capacity is generally not considered jurisdictional and is therefore waived if not specifically raised. See Fed. R.Civ.P. 9(a); 6A Charles Alan Wright et al., Federal Practice and Procedure § 1559, at 42 (2d ed.1990). However, if there is a possibility that a party with a valid capacity defense is strategically forgoing it to preserve federal jurisdiction, we believe it is proper for a federal court to assess the capacity of the parties sua sponte. See 28 U.S.C. § 1359 (“A district court shall not have jurisdiction of a civil action in which any party, by assignment or otherwise, has been improperly or collusively made or joined to invoke the jurisdiction of such court.”).
In this case, the parties assert that any Lloyd’s Name may be sued in his individual capacity directly by an insured. Some have suggested, however, that this may not be the case as a matter of British law (which would cover Haycock’s and Merrett’s capacity to be sued as individuals, assuming, as we do, that both are U.K. domiciliaries). See Indiana Gas, 141 F.3d at 321 (petition for rehearing) (stating, without citation, that English law does not permit such suits); cf. also Roby v. Corporation of Lloyd’s, 796 F.Supp. 103, 106 (S.D.N.Y.1992) (Lasker, J.) (quoting an English barrister’s affidavit indicating that the Names in a Lloyd’s syndicate cannot be bound by res judicata unless a suit is brought against all of the individual Names or against a representative Name). We take no position on the question of whether British law and the contracts in dispute in this case would permit a suit to proceed against a Name in his or her individual capacity. Instead, we include this among the issues to be resolved by the district court on remand.23 The following discussion of federal jurisdiction over an action brought against a Name in his individual capacity assumes without deciding that the district court will determine that such a suit is proper under Rule 17(b) and the contracts at issue here.
Whether it is possible to add Mer-rett in his individual capacity and subsequently to dismiss Haycock—either altogether or just in his representative capacity—and whether such steps, if feasible, would suffice to preserve jurisdiction in this case is something that the district court must, in the first instance, decide. But we do conclude that, if Haycock and Merrett properly remain in this suit only in their individual capacities, the existence of jurisdiction will depend solely on their citizenship—which is admittedly diverse—and on their meeting the requisite jurisdictional amount. It will not depend on the status of the other Lloyd’s Names who, though members of the syndicates at risk, would no longer be direct parties to the litigation.
The fact that a Lloyd’s policy contractually binds each of the underwriting Names to whatever judgments are rendered against any Name sued does not require us to consider the citizenship, of unspecified Names who are not before the court, so long as the only Name sued is properly in the litigation solely as an individual and not as a representative.24 Thus, we regularly exercise diversity jurisdiction in cases where non-diverse individuals or groups who are not direct parties to the litigation nevertheless have a crucial interest in its outcome. For example, jurisdiction exists when the diverse parties are, by contract, bound to be indemnified by non-diverse non-parties. See Wheeler v. City of Denver, 229 U.S. 342, 348, 350, 33 S.Ct. 842, 57 L.Ed. 1219 (1913) (holding that jurisdiction exists in such cases so long as the diverse parties exercise control over the litigation). And the fact that a case will almost certainly determine the rights of innumerable non-diverse litigants through collateral estoppel or preclusion in no way affects our jurisdiction if the party being-[937]*937directly sued is a real party to the controversy.25
There is no doubt that in their individual capacities Haycock and/or Merrett are real parties to this controversy because both are severally liable as Names on Squibb’s policy. It follows that, if the Lloyd’s Names represented by a lead underwriter are dispensable parties, see infra, then once these Names have been dropped from the suit, only the Name or Names who remain need to be completely diverse, given that they are themselves potentially liable and are being sued in their individual capacity.
In this respect it is easy to distinguish such a suit from a suit against a limited partnership. A limited partnership, like other unincorporated associations, is a formal entity. And, not by chance, the Supreme Court in Carden emphasized the importance of the existence of such an artificial entity in its holding that every limited partner must have diverse citizenship. See Carden, 494 U.S. at 187-88 n. 1, 110 S.Ct. 1015. Indeed, Carden’s express language applies only to “artificial entities created by state law.” See id. at 187, 110 S.Ct. 1015. There is, moreover, no language whatsoever in the Carden opinion suggesting that it was meant to govern situations where no formal entity is being sued. We therefore conclude that Carden only refers to situations in which a formal entity is a named party to a diversity action.
It might be argued, however, that Carden should be read to go beyond formal entities and also cover individuals bound by contract in such a way that they might be said to form a “constructive entity.” But even if Carden could be read to cover constructive entities created by contract, a Lloyd’s policy seemingly does not create such an entity. The contractual provision that obligates a Name to abide by a judgment rendered against any other Name runs vertically between the insured and each Name, not horizontally from Name to Name. As a result, the Lloyd’s policy is more aptly described as a series of independent bilateral contracts from insurer to insured than as a set of intertwining agreements uniting the Names in a manner analogous to a limited partnership. It is possible that horizontal contractual arrangements among the Names and the lead underwriters may create a constructive entity, but based just on the vertical contracts, a group of Lloyd’s Names is far less of an entity than groups bound by contract to each other to which Carden has never been thought to apply (e.g., the standard indemnity situation in which the indemnitor and indemnitee are contractually linked).
Nor does the fact that the individual Names hire a lead underwriter to be their agent for various administrative purposes alone suffice to create a constructive entity. A set of individuals may and often do share an insurance broker. But that has never been deemed relevant for jurisdictional purposes. It only becomes different when the agent is sued as a representative of these individuals. In that situation, the doctrine of Northern Trust controls and, as we said earlier, the citizenship of each person has to be completely diverse unless one of the three exceptions to the rule — corporations, trusts, or class actions — applies. See Northern Trust, 899 F.2d at 594.26
[938]*938In sum, the Lloyd’s market, because of its unusual legal arrangements: (a) can sometimes be covered by Northern Trust (e.g. whenever the lead underwriter is being sued in his representative capacity); but (b) can also be no more than an assortment of people whose only common ties are that they all may have legally independent contractual arrangements with the same insured and have hired the same agent (lead underwriter). If a Name is properly sued solely in his individual capacity to establish his individual liability, it is the latter relationship that defines the case for jurisdictional purposes. In such a case, that individual, with diverse citizenship, is before the court because he may be liable. And the fact that, if he is liable, others will—by contract, preclusion, or both—also have to pay is no more relevant than in the standard diversity torts case in which other, non-diverse, parties may, by contract of indemnity, by preclusion, or by both, ultimately have to answer for part of the judgment awarded.
. IV
Unfortunately, however, we are ill-suited to decide today whether Haycock can be dismissed in his representative capacity as a “dispensable party” pursuant to Newman-Green, New York law, and Rule 19 of the Federal Rules of Civil Procedure.27 The dis-pensibility of the Names represented by Haycock not only involves fact questions that we are not equipped to address, it also raises issues of law upon which there has been no briefing by the parties. The lack of briefing together with the factual questions on which an answer to the Rule 19 inquiry may depend induce us not to decide the issue on appeal but to remand it and various other related questions to the district court instead.
On remand,' the district court must pay due attention to Newman-Green and Caterpillar’s statements about the desirability of preserving jurisdiction. But it must also “carefully consider whether the dismissal of a nondiverse party will prejudice any of the parties in the litigation.” Newman-Green, 490 U.S. at 837-38, 109 S.Ct. 2218. As such, it must examine critically the claim of at least one Appellant (Continental Casualty Co.) that it will be prejudiced by a rechar-acterization of this suit. In this respect, it must also consider whether the dismissal of only one aspect of a single defendant’s dual participation in a suit—in this ease Haycock’s representative function as opposed to his individual role—is the equivalent of dismissing a “party” to the litigation under Newman-Green. And, if it finds that such partial dismissal presents particular problems, it must determine whether a total dismissal of Haycock is appropriate and feasible. But, most important, it must address, after briefing by the parties, the question of the indispensability, under Rule 19 and New York law, of the Names currently represented by Haycock.28
Furthermore, it must consider the questions (a) of whether Merrett can be added as a party defendant, and (b) of the jurisdictional amount to which he and Haycock are subject. In this respect, we see no special difficulty with adding Merrett as a party. See Fed.R.Civ.P. 21 (“Parties may be [939]*939dropped or added by order of the court on motion of any party or of its own initiative at any stage of the action and on such terms as are just”). But it is far from clear whether Merrett must meet: (1) the $10,000 amount in controversy mandated when the suit was begun, see 28 U.S.C. § 1332(a) (1982), amended by Pub.L. No. 100-702, Tit. 11 § 201(a), 102 Stat. 4646 (1988) (increasing the requisite amount to $50,000), and Pub.L. No. 104-317, Tit. 11, § 205(a), 110 Stat. 3850 (1996) (raising the amount to $75,000); (2) the current amount in controversy of $75,000, see 28 U.S.C. § 1332(a) (Supp.1998); or (3) no amount in controversy at all, see 28 U.S.C. § 1332(a)(3). Unless, moreover, he meets whatever requirement applies to him, adding him would be futile and the motion to do so should be denied.
We are skeptical that § 1332(a)(3) can be read so broadly as to eliminate entirely the amount-in-controversy requirement for foreign subjects. But that issue — like the question of whether the relevant amount is that which governed when the suit was brought or is the current amount — will be mooted if Merrett meets the $75,000 amount in controversy now required. Accordingly, we decline to decide them. Should the district court find that they must be addressed in order to determine whether Merrett ought to be added, it can deal with them then.
V
Accordingly, the district court, on remand, should first consider whether Merrett meets the jurisdictional requirements and, hence, whether he should in the ordinary course be added as a party defendant. Next, the court should determine whether the suit can proceed without a defendant who formally represents the Lloyd’s Names, that is, whether Squibb can sue Merrett and/or Haycock only in their individual capacities.29 For such individual suits to proceed, the district court must find: (1) that such a suit is permitted as a matter of both British law and the contracts of this ease;30 (2) that the previously represented Lloyd’s Names are dispensable parties under Rule 19 and New York law and therefore can be dropped; (3) that no undue prejudice to any defendant is created by recharacterizing the suit as one that is brought solely against an individual Lloyd’s Name; (4) that Haycock can, under New-marir-Green, be dropped in his representative capacity only; or (5) if Haycock must be totally dismissed, that (after the addition of Merrett) Haycock is a dispensable party and hence can be excused from the case.
If a suit against Merrett and/or Haycock as an individual is not possible, then and only then should the district court consider whether Squibb’s. complaint ought to be recast as a Rule 23.2 class action against the Lloyd’s underwriters. Before recharacteriz-ing this suit as a Rule 23.2 action, the district court must determine whether the Lloyd’s underwriters constitute a proper Rule 23.2 class and, if they do, whether and by whom the amount-in-controversy requirement must be met. If each and every Name does not meet the amount in controversy, the court must then decide: (1) whether 28 U.S.C. § 1367 has overruled Zahn retroactively and permits aggregation in class actions of this sort; (2) whether aggregation is permitted nonetheless in a Rule 23.2 class action; or (3) whether 28 U.S.C. § 1332(a)(3) has abolished the amount-in-controversy requirement for foreign subjects, leaving only the citizen Names subject to the requirement.
VI
We hold that when a Lloyd’s lead underwriter is sued in a representative capacity (but not in a class action) each and every Name whom the lead underwriter represents must be completely diverse. But we also hold that when a Lloyd’s Name (including a lead underwriter) is properly sued only in an individual capacity, it is that Name’s characteristics, both as to citizenship and jurisdictional amount, that are determinative for jurisdictional purposes. And the fact that other [940]*940Lloyd’s underwriters who are not diverse parties in the suit may be bound by the result of the suit (whether by contract or by preclusion) is of no consequence. All other issues in this case we decline to decide and remand to the district court.
Although we are vacating and remanding after sixteen years of federal litigation, we do so reluctantly and in the spirit of Newman-Green in an attempt to salvage federal jurisdiction, if jurisdiction can be saved by means of the alternatives we have presented and within the limits set by the relevant statutes and holdings of the Supreme Court.
The judgment of the district court is vacated and the ease is remanded for further proceedings consistent with this opinion.31 This panel will retain jurisdiction — such as it is — over any subsequent appeals.