In May 1988, Donald Johnson brought suit against Patrick Burken in an Illinois state court for damages arising out of an automobile accident that had occurred in August 1986. At the time of the accident Burken was a sailor stationed at the Great Lakes naval base in Illinois, although he was a domiciliary--and hence, for purposes of federal diversity jurisdiction, a citizen, Galva Foundry Co. v. Heiden, 924 F.2d 729 (7th Cir.1991)--of Iowa. But by the time the suit was filed, Burken had left Illinois for duty at sea. Illinois' nonresident motorist statute allows service of process on a nonresident by serving a copy of the complaint or other process on the secretary of state of Illinois and, within ten days thereafter, sending a copy "by registered mail" to the defendant at his last known address. Ill.Rev.Stat. ch. 95 1/2, p 10-301(b). Johnson's lawyer duly served the secretary of state with his copy, but mailed Burken's copy by certified mail (return receipt requested) rather than by registered mail. Burken did not receive his copy, having failed to leave a forwarding address.

Several months later, Johnson's lawyer wrote Burken's insurer suggesting settlement. This was the first that either Burken or the insurance company knew of the suit. The insurance company retained a lawyer to defend the case. He made a special appearance to contest jurisdiction and moved that the case be dismissed on the ground that service had not been made in accordance with the statute and therefore the court had not obtained jurisdiction over Burken. The state judge denied the motion without giving reasons. Burken then removed the case to federal district court--the parties being of diverse citizenship--and renewed his motion to dismiss for lack of proper service. In September the district judge denied the motion on the ground that the state judge's ruling had not been clearly erroneous and therefore was binding in subsequent proceedings under the doctrine of the law of the case. Burken then moved for a stay of proceedings under the Soldiers' and Sailors' Civil Relief Act, 50 U.S.C.App. Secs. 501 et seq. (see Sec. 521), and also moved for a certification to permit him to take an immediate appeal to this court under 28 U.S.C. Sec. 1292(b). In January of last year the district judge granted both motions. Shortly before, in November 1989, Johnson had again attempted to serve Burken under the Illinois nonresident statute, this time sending him his copy in Norfolk, Virginia, where Burken was stationed. The validity of this service has not been determined; nor whether it was made within the statute of limitations. A motions panel of this court agreed to hear Burken's interlocutory appeal.

Three jurisdictional issues require consideration before we can get to the merits of the appeal. The first is whether the stay that the district judge granted under the Soldiers' and Sailors' Civil Relief Act bars further proceedings in this court as well as in the district court. The Act entitles a member of the armed forces to a stay of legal proceedings to which he is a party while he is on active duty, unless the judge decides that the serviceman's litigation prospects are not materially affected by his service status. Nothing in the Act excepts proceedings in a Court of Appeals, Kesler v. Kesler, 682 S.W.2d 44, 45 n. 1 (Mo.App.1984)--on the contrary, the Act applies "at any stage" of a case, although it is rare that being on active duty will "materially affect" a litigant's case on appeal, since he does not have to be present in person. Id.; Briner v. Briner, 60 Cal. App. 2d 473, 140 P.2d 995 (1943). But none of this is important, because the Act does not limit the jurisdiction of the federal courts. It merely gives a procedural right to a class of litigants. Like most procedural rights, it can be waived. Roberts v. Fuhr, 523 So. 2d 20, 28 (Miss.1987); McMahon v. McMahon, 70 Cal. App. 2d 317, 160 P.2d 893 (1945); Lankford v. Milhollin, 201 Ga. 594, 600-01, 40 S.E.2d 376, 380 (1946); Deacon v. Witham, 131 Misc. 2d 217, 499 N.Y.S.2d 317 (City Ct.1985). In asking for leave to take an interlocutory appeal, Burken waived his right to a stay of proceedings in this court--he could not appeal and stay the appeal at the same time. Alternatively and with the same effect, he sought and received merely a limited stay. Nothing in the Act precludes a limited stay.

The second jurisdictional question is whether the appeal is proper under section 1292(b). We are free to reexamine the motion panel's determination of that question unembarrassed by the law of the case doctrine or by the requirement that a panel opinion overruling a decision of this court be circulated in advance to all the judges in regular active service. 7th Cir.R. 40(f). Decisions by motions panels are summary in character, made often on a scanty record, and not entitled to the weight of a decision made after plenary submission. United States v. Houser, 804 F.2d 565, 568 (9th Cir.1986). Certainly when the panel is merely deciding whether an appeal should be heard, rather than disposing of the appeal, its decision should be regarded as tentative, and therefore revisable by the merits panel.

Section 1292(b) requires that the ruling sought to be appealed "involves a controlling question of law as to which there is substantial ground for difference of opinion and that an immediate appeal from the order may materially advance the ultimate termination of the litigation." Whether strict or substantial compliance is the test under the Illinois nonresident motorist statute is a question of law, and being a threshold question its resolution now may materially advance the ultimate termination of the litigation. But is it a "controlling" issue, given that the validity of the second attempt at service has not been determined? If the second service was completed within the period of limitations and was valid, the validity of the first service is immaterial.

We think "controlling" means controlling in light of the situation that exists when the district judge makes his certification under section 1292(b) and when we exercise our discretion to accept the certified ruling for an immediate appeal. True, the fact that the litigation may later take a twist or turn that demotes the "controlling" question's significance, or that the district judge's ruling on the question may later be discovered not to have been dispositive, would permit us to dismiss the appeal as having been improvidently granted. Molybdenum Corp. of America v. Kasey, 279 F.2d 216 (9th Cir.1960) (per curiam); 16 Charles A. Wright, Arthur R. Miller, Edward H. Cooper & Eugene Gressman, Federal Practice and Procedure Sec. 3929, at p. 142 (1977). But it would not require us to do so, as would a development that made the district judge's ruling of which an immediate appeal was sought moot. Moreau v. Tonry, 554 F.2d 163 (5th Cir.1977) (per curiam); see also New York City Health & Hospitals Corp. v. Blum, 678 F.2d 392, 397 (2d Cir.1982); Oneida Indian Nation v. County of Oneida, 622 F.2d 624, 628-29 (2d Cir.1980). Jurisdiction ordinarily is determined by the facts that exist when the case is filed, Newman-Green, Inc. v. Alfonzo-Larrain, 490 U.S. 826, 830, 109 S. Ct. 2218, 2222, 104 L. Ed. 2d 893 (1989); Smith v. Sperling, 354 U.S. 91, 93 n. 1, 77 S. Ct. 1112, 1113 n. 1, 1 L. Ed. 2d 1205 (1957); appellate jurisdiction, therefore, by the facts that exist when the appeal is filed. Of course a court can lose jurisdiction, as when a case becomes moot, but the fact that subsequent events in a litigation made a ruling less plainly dispositive than it had seemed at first would not do this. One reason that events subsequent to filing ordinarily do not destroy jurisdiction is that, if they did, this would be an invitation to strategic behavior. For example, a party might change his state of citizenship in order to defeat diversity jurisdiction--might do it even after judgment had been rendered against him in the diversity suit. Another reason, more pertinent here, for disregarding postfiling developments is that, otherwise, piecemeal litigation would be encouraged. Imagine if the plaintiff had made his second attempt at service after the filing of the appeal from the district judge's ruling concerning the first attempt and had then moved to dismiss the appeal on the ground that it no longer involved a controlling question of law.

Yet such a tactic would succeed if "controlling" were read literally. It is not read literally. It could not be, because it is never one hundred percent certain in advance that the resolution of a particular question will determine the outcome or even the future course of the litigation. Therefore "a growing number of decisions have accepted the rule that a question is controlling, even though its decision might not lead to reversal on appeal, if interlocutory reversal might save time for the district court, and time and expense for the litigants." 16 Wright, Miller, Cooper & Gressman, supra, Sec. 3930, at pp. 159-60 (footnote omitted). Illustrative decisions are In re Cement Antitrust Litigation, 673 F.2d 1020, 1026-27 (9th Cir.1982), and Katz v. Carte Blanche Corp., 496 F.2d 747, 755 (3d Cir.1974), the latter stating that " 'controlling' means serious to the conduct of the litigation, either practically or legally."

Practical considerations well illustrated by this case show that the standard should be kept flexible. If we don't decide the validity of the first service, the case may go through to judgment, followed by an appeal that will result (as we are about to see) in throwing the case out for want of proper service--thus requiring a remand for determination of the validity of the second attempt at service, followed possibly by another appeal. Even so, we might have been inclined to dismiss the appeal had it not been briefed and argued before the potential significance of the second service was brought forcefully to our attention. But since it has been briefed and argued, we think it best to decide the appeal and speed the litigation on its way rather than dismiss the appeal and by doing so invite the potential consequences sketched above. There is no danger that a flexible standard will result in inundating the courts of appeals with appeals under section 1292(b), since our permission to take such an appeal is required and may be withheld in our discretion, without even a statement of reasons. United States v. Woodbury, 263 F.2d 784, 788 (9th Cir.1959).

The third jurisdictional issue is whether the case was removed to the district court in timely fashion. It was not, 28 U.S.C. Sec. 1446(b); Lewis v. Louisville & Nashville R.R., 758 F.2d 219, 222 (7th Cir.1985), but that is immaterial because the plaintiff failed to object to removal until the appeal was filed. That was much too late. A removed case normally and in the present instance is a case within the original jurisdiction of the federal district court. If the plaintiff does not object to the removal it is as if he had refiled the case in the federal district court, as in Illinois ex rel. Barra v. Archer Daniels Midland Co., 704 F.2d 935, 939 (7th Cir.1983). He cannot be permitted to change his mind later and toss the case back to state court. Nolan v. Prime Tanning Co., 871 F.2d 76 (8th Cir.1989); Meadows v. Bicrodyne Corp., 785 F.2d 670, 672 (9th Cir.1986).

We come at last to the merits. The Illinois nonresident motorist statute requires service at the defendant's last-known address "by registered mail." Certified mail is not registered mail. But when a return receipt is requested by the certified mailer, the only difference is that the Postal Service keeps a record of registered mail, thereby facilitating the reconstruction of the exact course and timetable of the missive. This difference may seem slight. In Olin Corp. v. Bowling, 95 Ill.App.3d 1113, 51 Ill. Dec. 489, 420 N.E.2d 1047 (1981), the Illinois Appellate Court allowed the plaintiff in a class action to use certified mail to serve the 3,801 members of the defendant class in order to save the plaintiff the added cost of registered mail (an extra $2.55 per letter, at the time). But however elastic the statute may be to interpretation, it cannot be stretched to embrace a garden-variety case of attorney error, which is all that is involved in this case. For then we would simply have substituted "certified" for "registered" mail in the statute, since every rational plaintiff's lawyer would use the cheaper and less certain form of mailing if he were permitted to do so. We are not devotees of literal interpretation; but only a court with a power of statutory revision could equate "certified mail" to "registered mail" in a case in which (unlike Olin Corp.) there was no possible argument that the legislature had failed to anticipate a situation in which literal compliance with the statute would be unduly, perhaps senselessly, burdensome.

The doctrine of law of the case does not require us to accept the state trial judge's contrary interpretation. There is some confusion about the doctrine, which we shall take this opportunity to try to dispel. Originally it just meant that a lower court is bound by the rulings of the higher court made in the same case. Sibbald v. United States, 37 U.S. (12 Pet.) 488, 492, 9 L. Ed. 1167 (1838). This is not a very startling principle. If the rulings by the higher court are made in a different case, the lower court is bound by one of the less controversial aspects of the doctrine of stare decisis. A fortiori it is bound by its judicial superiors' ruling in the same case.

Controversy over the doctrine of law of the case properly focuses on its invocation by a judge asked to change a previous ruling of his in a case, or by judges asked to change a previous ruling by a coordinate (as distinct from superior) court in a case. In these, the only interesting applications of the doctrine, it is a doctrine about reconsideration. That is how it normally is expressed. Here is a typical formulation: "a court will ordinarily not reconsider its own decision made at an earlier stage of the trial or on a prior appeal, absent clear and convincing reasons to reexamine the prior ruling." Gertz v. Robert Welch, Inc., 680 F.2d 527, 532 (7th Cir.1982). It is important to give full weight to the word "decision." No one thinks that if a trial judge overrules an objection to the admission of some piece of evidence, and a moment later the lawyer against whom he ruled asks him to reexamine his ruling, the judge may not do so unless he has "clear and convincing reasons." No judge is so rigid. Judges are constantly reexamining their prior rulings in a case on the basis of new information or argument, or just fresh thoughts, without excogitating a "clear and convincing" reason for their change of heart. On the other hand, once a case has been decided, then, unless the decision was avowedly tentative (for example, a decision granting or denying a preliminary injunction or--our earlier example--a decision by a motions panel), there is a natural and healthy reluctance not to reconsider the decision unless powerful reasons are given for doing so. Otherwise parties would have an incentive constantly to pester judges with requests for reconsideration.

But all this is to one side of our case. Burken is not asking us to reconsider a previous decision of ours in this litigation; we have not made a previous decision. We may assume that for purposes of applying the doctrine the proceedings in the state court were "an earlier stage" of this litigation rather than a different case. "[T]he doctrine applies as much to the decisions of a coordinate court in the same case as to a court's own decisions," Christianson v. Colt Industries Operating Corp., 486 U.S. 800, 816, 108 S. Ct. 2166, 2177, 100 L. Ed. 2d 811 (1988), a precept applied to a removed proceeding in Barrett v. Baylor, 457 F.2d 119, 124 (7th Cir.1972). This extension is necessary to discourage strategically motivated removals and other forms of judge or court shopping. So the district judge was entitled not to reexamine the state judge's ruling. But Burken was equally entitled to appellate review of that ruling.

The following example may make the point clearer. Suppose that at the outset of a lawsuit the defendant moves to dismiss the suit on the ground that the statute of limitations has expired. The trial judge denies the motion. The denial is an interlocutory order and therefore not appealable at the time made. The case goes to trial and the plaintiff wins. The defendant moves for judgment notwithstanding the verdict, reasserting his statute of limitations defense. The judge invokes the law of the case doctrine and refuses to reexamine his earlier ruling. This does not prevent the defendant from arguing on appeal that that ruling was incorrect. One reason the ruling was not appealable when made was that it was appealable at the end of the case unless it became moot in the interim.

The order of the district judge is vacated and the case remanded for further proceedings consistent with this opinion.

RIPPLE, Circuit Judge, dissenting.

I would vacate the order granting leave to appeal on the ground that the order was entered improvidently.

The majority concedes that the validity of the first service may be "immaterial" in light of the second service. Ante at 1205. Thus, as the majority also recognizes, the question on which this section 1292(b) interlocutory appeal was certified may not be dispositive--or, in the word of the statute, "controlling." Nonetheless, the majority holds that, if this court discovers only after accepting jurisdiction that the question is not controlling, we are permitted, but not required, to vacate the appeal as having been improvidently granted. Id. at 1205-06. I have a different view of our responsibilities under section 1292(b).

The majority, relying on cases dealing with other statutes regulating our appellate jurisdiction, notes that the existence of the prerequisites to jurisdiction generally ought to be evaluated when the case is filed. Section 1292(b) is, however, a unique statute. Contrary to the general scheme of federal appellate jurisdiction, it permits interlocutory appeals. Notably, such appeals are permitted only when both the district court and the court of appeals are convinced that the issue certified meets the statutory criteria. In short, this court, no less than the district court, has the responsibility to determine that any appeal taken under this statute is compatible with the congressional intent to abrogate the general rule of finality only in the most narrow of circumstances.

The majority correctly notes that mootness, or any other problem rendering the case nonjusticiable in the constitutional sense, would prevent our deciding an issue certified to us by a district court. However, the absence of such a problem hardly leaves us with unfettered discretion to accept an appeal under section 1292(b). We still must determine whether proceeding to decision would be compatible with the purpose of the statute. Courts of appeals have long recognized that responsibility.1

The majority appears to recognize that we are bound to heed this congressional mandate when it concedes that it "might have been inclined to dismiss the appeal had it not been briefed and argued before the potential significance of the second service was brought forcefully to our attention." Id. at 1206.2 In fact, the motions panel was informed about the second service of process before it granted permission to proceed with this interlocutory appeal. On the first page of Plaintiff Johnson's Answer to Petition for Permission to Appeal Interlocutory Order, he stated that the defendant "was properly served again (by registered mail) in November of 1989, making this issue moot." Thus, that panel was on notice that the validity of the first service might no longer be a controlling question in this case.3

In this case, another factor counsels extreme circumspection before we permit an interlocutory appeal. Prior to the removal of this case to federal court, a state court, in the normal course of litigation, had ruled on the issue of the validity of service of process through certified rather than registered mail. While we are not bound by the law of the case doctrine, such matters of local state practice seem particularly inappropriate for discretionary review. Whether the plaintiff's use of certified mail amounted to substantial compliance with the statute is a matter, given considerations of federalism, that we ought not address unnecessarily. With our present heavy caseload, this otherwise routine state law tort suit should have been resolved by further proceedings in the district court rather than interrupted by an interlocutory appeal. I respectfully dissent.