dissenting from the denial of rehearing en banc.
A class action contending that the Bank of Boston and its affiliates (collectively, the Bank) did not promptly post interest to real estate escrow accounts was filed in Alabama by a Chicago law firm. Hoffman v. BancBoston Mortgage Corp., No. CV-91-1880 (Mobile Cty. Cir. Ct.). Settlement ensued, and the class members learned only what the notice told them. Few opted out or objected, because the maximum award to any class member was less than $9. Any recovery, however small, seemed preferable to initiating a separate suit or even bearing the costs of protesting the settlement’s terms. After the state judge approved the pact, the Bank carried out its part: it disbursed more than $8 million to the class attorneys in legal fees and credited most accounts with paltry sums. Problem: the fees, equal to 5.32 percent of the balance in each account, were debited to the accounts. For many accounts the debit exceeded the credit. Dexter J. Kamilewicz, for example, received a credit of $2.19 and a debit of $91.33, for a loss of $89.14.
Outraged account holders hired new lawyers to sue their “champions” from the state case for malpractice, breach of fiduciary duty, indeed for fraud. Defendants insist that most members of the class emerged with net benefits, but they do not deny that many suffered net losses — and that these impending losses, known to those who negotiated the settlement, were not disclosed to the class members. Kamilewicz believes that even when the credit exceeded the debit the account holder lost, because the credit would have been made when the mortgage terminated and the escrowed sum was released; the real economic benefit was not the amount of the credit but the interest that could be earned from investing the payment sooner.
Kamilewicz and the other representative plaintiffs sued not only the lawyers for the plaintiff class in Hoffman but also the Bank and the Bank’s lawyers. Everything the Bank did to affect Kamilewicz’s interests was authorized by the judgment based on the settlement, so the Bank’s inclusion in the case marked an effort to obtain collateral review of the state court’s judgment. Kami-lewicz argued that the judgment is void with respect to account holders who live outside Alabama — for what right does Alabama have to instruct financial institutions in Florida to debit the accounts of citizens of Maine and other states? (Despite its name, Banc-Boston’s headquarters are in Florida, and the Kamilewiezes hail from Maine.) Class members are not bound when a state court exceeds its jurisdiction. The Attorneys General of Florida, Maine, and seven other states have filed a brief as amicus curiae to protest their citizens’ treatment by Alabama. According to Kamilewicz, the Alabama judgment is doubly flawed, because the notice was so misleading that it denied the class members due process of law. A deficient notice means that the class members’ right to opt out was defeated, which has jurisdictional consequences of its own. Lack of jurisdiction and unconstitutional procedures are recognized grounds for disregarding a judgment in a class action. Phillips Petroleum Co. v. Shutts, 472 U.S. 797, 105 S.Ct. 2965, 86 L.Ed.2d 628 (1985); Richards v. Jefferson County, — U.S. -, 116 S.Ct. 1761, 135 L.Ed.2d 76 (1996). But by asking for collateral review of a judgment, the Kamilewicz class invited the response: “In federal court?” Under the Rooker-Feldman doctrine, named after Rooker v. Fidelity Trust Co., 263 U.S. 413, 44 S.Ct. 149, 68 L.Ed. 362 (1923), and District of Columbia Court of *1350Appeals v. Feldman, 460 U.S. 462, 103 S.Ct. 1303, 75 L.Ed.2d 206 (1983), the Supreme Court of the United States is the only federal court authorized to review a state court’s judgment in civil litigation. Defendants invoked this rule and argued that the Kamile-wicz class could obtain relief only from a state court. The district court agreed and dismissed the entire complaint for want of jurisdiction, and a panel of this court affirmed. 92 F.3d 506 (1996).
Suppose A sues B in a state that lacks any connection to B or to the dispute. B refuses to appear and the court enters a default judgment, which A attempts to collect using the processes of the federal court under the diversity jurisdiction. Would the court really direct B to pay, without inquiring into the jurisdiction of the state court? Certainly not in the fifth circuit, which has held that the .Rooker-Feldman doctrine permits the kinds of collateral attacks that any state court of general jurisdiction could entertain. Davis v. Bayless, 70 F.3d 367, 375-76 (5th Cir.1995). (“Any” is an important qualifier, for reasons discussed in Homola v. McNamara, 59 F.3d 647, 650 (7th Cir.1995), and Harris Trust & Savings Bank v. Ellis, 810 F.2d 700 (7th Cir.1987).) When Hansberry v. Lee, 311 U.S. 32, 61 S.Ct. 115, 85 L.Ed. 22 (1940), held that a judgment rendered without personal jurisdiction over members of the class is not entitled to full faith and credit, the Justices did not add: “except in federal court.” Granted, the defendants in today’s ease are not trying to execute a judgment; they want to fend off a challenge to the status quo. But the difference between offensive and defensive use of a judgment is not important to the panel’s conclusion, or to mine — which is that a judgment that is not entitled to full faith and credit does not acquire extra force via the Rooker-Feldman doctrine.
Collateral attacks based on lack of personal or subject-matter jurisdiction are proper, no less in class actions than in other cases— indeed, they are especially appropriate where class members are stunned to find that, although aligned as plaintiffs, they are net losers, just as if the original defendants had filed and prevailed on a counterclaim of which they received no notice and over which the state court had no jurisdiction. In effect, though not in name, this was a defendant class, attempting (unbeknownst to its members) to fend off predatory lawyers’ claims to the balances in the escrow accounts. The substantial jurisdictional problems entailed by defendant classes, see Phillips Petroleum, 472 U.S. at 808, 811 & n. 3,105 S.Ct. at 2972-73, 2974 & n. 3; Diane P. Wood, Adjudicatory Jurisdiction and Class Actions, 62 Ind. L.J. 597, 607-12 (1987), were aggravated by the class members’ ignorance of their exposure.
In Matsushita Electric Industrial Co. v. Epstein, — U.S. -, 116 S.Ct. 873, 134 L.Ed.2d 6 (1996), members of a class in a securities case filed in state court initiated an independent action in federal court for the purpose of attacking the terms on which the state case had been settled. They argued that the state court lacked jurisdiction to resolve claims based on federal securities law. Ultimately they lost — but not on Rook-er-Feldman grounds. The Supreme Court concluded that state judges may accept and enforce settlements of federal claims. This it held on the merits. It did not mention the possibility that the federal court lacked jurisdiction, which I think signifies how obvious it was to all concerned that the Rooker-Feld-man doctrine is inapplicable. If it was inapplicable in Matsushita, it is inapplicable here too. I grant that Matsushita did not “hold” this, because it did not discuss jurisdiction. Pennhurst State School & Hospital v. Halderman, 465 U.S. 89, 119 & n. 29, 104 S.Ct. 900, 918 & n. 29, 79 L.Ed.2d 67 (1984); United States v. L.A Tucker Truck Lines, Inc., 344 U.S. 33, 37-38 & n. 9, 73 S.Ct. 67, 69 & n. 9, 97 L.Ed. 54 (1952). Shared assumptions are revealing nonetheless. And in today’s case as in Matsushita the aggrieved class members may choose a new forum; they need not return to a court whose lack of jurisdiction is the basis of their position. In re Real Estate Title & Settlement Services Antitrust Litigation, 869 F.2d 760, 767-69 (3d Cir.1989). None of this is to say that the Alabama court actually lacked jurisdiction, or that the notice was inadequate; to get anywhere against the Bank the Kamilewicz class must establish these things; but under the panel’s approach the state court’s judgment *1351blocks federal litigation even if it lacked jurisdiction.
Next consider plaintiffs’ claim against the Hoffman class counsel, which is not a collateral attack on a judgment. It takes the judgment as a given — indeed, it is only so long as the judgment stands that the litigant has a compensable loss. Neither state nor federal law requires a malpractice suit to be filed in the same court that handled the initial litigation. The Rooker-Feldman doctrine therefore does not apply to malpractice suits, which may be litigated in federal courts without regard to the location of the initial case. If the panel is right, no malpractice suit growing out of state litigation in which the judge awarded attorneys’ fees — maybe no malpractice suit, period — may be brought in federal court, even if all requirements of the diversity jurisdiction have been satisfied. This holding is sufficiently troubling and affects so many other cases that it is worth the time of our court to consider the subject en bane.
The Rooker-Feldman doctrine has two rationales: first, that the decision of a state court reached after notice and opportunity for hearing is not an independent violation of the Constitution, and therefore is not actionable under 42 U.S.C. § 1983; second, that by virtue of 28 U.S.C. § 1257 only the Supreme Court may review the decision of a state court in civil litigation. See ASARCO Inc. v. Kadish, 490 U.S. 605, 622-23, 109 S.Ct. 2037, 2048-49, 104 L.Ed.2d 696 (1989); GASH Associates v. Rosemont, 995 F.2d 726 (7th Cir.1993). The rationales of the doctrine identify important limits: the first theme is irrelevant to a malpractice suit, and the second is beside the point, because the plaintiff does not seek review of the decision. The original parties’ entitlements vis-a-vis each other are fixed by the judgment. The contest now is between the litigant and his lawyer. We have emphasized this distinction many times — not only in malpractice eases, but also when the loser tries to get relief from judgment because of the lawyer’s incompetence. E.g., United States v. 7108 West Grand Avenue, 15 F.3d 632 (7th Cir.1994).
Consider a related sequence of litigation. A sells a painting to B, and before delivering the painting to B sells it again to C. Then A learns that the painting has increased in value and holds onto it. B sues A in state court for specific performance and prevails. Next C sues A for damages. Surely the. Rooker-Feldman doctrine does not foreclose C from using federal court. C’s suit does not attack the judgment (B keeps the painting); C wasn’t even a party to B v. A; and the cause of C’s loss wasn’t the judgment, but A’s double dealing. So too with malpractice litigation: it is a suit against a nonparty (the lawyer) alleging harm from incompetent or deceitful acts. That the lawyer’s misconduct occurred in a judicial proceeding doesn’t insulate the lawyer irom liability, even when the Rooker-Feldman doctrine insulates the judgment.
The Supreme Court’s most recent discussion of the Rooker-Feldman doctrine strongly implies that it does not bar malpractice litigation. Johnson v. De Grandy, 512 U.S. 997, -, 114 S.Ct. 2647, 2654, 129 L.Ed.2d 775 (1994), holds that the Rooker-Feldman doctrine does not bar voting rights litigation filed by the United States, even though the districts being challenged were approved by a state court in adversarial litigation. The Court observed that the United States was not a party to the state case and therefore could not have presented its claim by petition for certiorari under § 1257; the inability to do so meant that it could litigate independently. We, too, have held that the Rooker-Feldman doctrine does not affect suits by or against persons who were not parties to the initial case. Leaf v. Supreme Court of Wisconsin, 979 F.2d 589, 598 (7th Cir.1992); Downs v. Westphal, 87 F.3d 202, 203 (7th Cir.1996) (Rovner, J., concurring). Cf. Lynk v. LaPorte Superior Court No. 2, 789 F.2d 554 (7th Cir.1986). It may apply when the non-parties occupy positions functionally identical to the parties (for example, when a person whose property has been condemned by a city retaliates with a suit against the mayor), but the class attorneys are not the Bank’s alter egos. The attorneys representing the Hoffman class were not parties to the Alabama ease. Neither were the class members. Absent class members are represented by the named plaintiffs and their lawyers, *1352but they aren’t parties, a point reflected in federal litigation by disregarding their citizenship. They are ignored in negotiating settlements as well. A real party’s lack of assent means that there is no settlement; but the missing class members don’t sign the settlement, and their objection is not disposi-tive. It is crammed down the throats of objectors, which cannot be done to real parties. What is more, the members of the Kamilewicz class could not have presented their malpractice arguments to the Supreme Court by petition for certiorari under § 1257: they did not discover the malpractice until later (it was not reflected in the record of the state proceeding); it was not litigated in the Hoffman case; class members can’t seek appellate review without intervening, In re VMS Limited Partnership, 976 F.2d 362 (7th Cir.1992); Research Corp. v. Asgrow Seed Co., 425 F.2d 1059 (7th Cir.1970), which further illustrates their non-party status; and of course malpractice is not a federal claim, so it is outside the scope of § 1257. For some purposes missing class members are treated like parties, but only if the named plaintiffs adequately represent the interests of the class, and only if the unnamed members of the class receive adequate notice and elect not to opt out, Phillips Petroleum, 472 U.S. at 808-13, 105 S.Ct. at 2972-75—which in this ease is the very thing in dispute! It gets the cart before the horse to reject, as barred by a judgment, an effort by the absent class members to show that they were not properly brought into the state ease and therefore are not affected by the judgment.
From all of this it follows that a malpractice action is not affected by the Rooker-Feldman doctrine. Does the fairness hearing required to approve the settlement of a class action make a difference? I think not. For the reasons just explained, absent class members (especially those who deny the state court’s jurisdiction over them) are not parties and cannot be treated as bound by the findings implicit in the approval of the settlement and the award of fees to attorneys. The panel implied approval of the district court’s conclusion that the class members were effectively defendants resisting financial demands by the attorneys, 92 F.3d at 511, which makes the jurisdictional shortcoming clear and establishes that the judgment is not entitled to full faith and credit.
All jurisdictional doubts to one side, a settlement followed by a fairness hearing remains more like a contract than like litigation. Accordingly there is even less reason to apply the Rooker-Feldman doctrine than in a normal malpractice case, where the loss ensues from a genuine contest. Representative plaintiffs and their lawyers may be imperfect agents of the other class members— may even put one over on the court, in a staged performance. The lawyers support the settlement to get fees; the defendants support it to evade liability; the court can’t vindicate the class’s rights because the friendly presentation means that it lacks essential information. This possibility, a staple of the literature about class actions, e.g., John C. Coffee, Jr., Class Wars: The Dilemma of the Mass Tort Class Action, 95 Colum. L.Rev. 1343 (1995); Jonathan R. Maeey & Geoffrey P. Miller, The Plaintiffs’ Attorney's Role in Class Action and Derivative Litigation: Economic Analysis and Recommended tions. for Reform, 58 U. Chi. L.Rev. 1 (1991), enjoys judicial recognition. See Georgine v. Amchem Products, Inc., 83 F.3d 610, 630-31 (3d Cir.1996), cert. granted under the name Amchem Products, Inc. v. Windsor, — U.S. -, 117 S.Ct. 379, 136 L.Ed.2d 297 (1996); In re General Motors Corp. Pick-Up Truck Fuel Tank Products Liability Litigation, 55 F.3d 768 (3d Cir.1995). Recall that even the active opposition of class members does not automatically block the settlement, which therefore rests on the lawyers’ consent, not the litigants’ agreement or the legal decision of a court.
The Kamilewicz class asserts that it suffered harm from the Hoffman class lawyers’ breach of their duties of care and loyalty in negotiating the settlement, which was concealed from the Alabama judge (and the class) by a further breach of the duty of loyalty in drafting the notice about the settlement. The notice not only didn’t alert the absent class members to the impending loss but also pulled the wool over the state judge’s eyes. Suing faithless agents is far from the core of the Rooker-Feldman doc*1353trine, which should not be extended to block suits like this. “Were [plaintiff] merely claiming that the decision of the state court was incorrect, even that it denied him some constitutional right, the doctrine would indeed bar his claim. But if he claims, as he does, that people involved in the decision violated some independent right of his ... then he can, without being blocked by the Rooker-Feldman doctrine, sue to. vindicate that right.” Nesses v. Shepard, 68 F.3d 1003, 1004 (7th Cir.1995).
The panel’s decision is important. If I am right that the state court’s award of attorneys’ fees does not matter, then the decision logically bars all malpractice and related fiduciary-duty suits arising out of state litigation. We have entertained many, without seeing jurisdictional problems. And even if I am wrong about the sweep of the panel’s decision, the issue is recurrent. Two recent decisions of other circuits—Durkin v. Shea & Gould, 92 F.3d 1510 (9th Cir.1996), and Diaz v. Sheppard, 85 F.3d 1502 (11th Cir.1996)—hold that a class may prosecute a malpractice action against its attorneys even after a settlement has been approved in a fairness hearing. Although the initial suits in both Durkin and Diaz were in federal court, making the Rooker-Feldman doctrine inapplicable on our panel’s holding, the approach of Durkin and Diaz is incompatible with the panel’s. If the Rooker-Feldman doctrine applies to suits by the absent class members because a malpractice action is a collateral attack on the order .approving the settlement and awarding attorneys’ fees, then the law of preclusion (res judicata) should bar malpractice actions in any court, state or federal, and without regard to which judicial system handled the first case. Yet no one thinks that. A malpractice suit is an independent action. A (potential) defense of issue preclusion is defeated by the very theory of the claim: that the first judgment is unreliable because of the attorney’s bungling. The bungler cannot point to the adverse, judgment produced by his own incompetence to ward off the client’s demand. The Kamilewicz class may fail in its proof, or it may encounter other obstacles, but the Rooker-Feldman doctrine does not close the door of the federal courthouse.