Daniel Sims fell from a lift platform made by EGA Products and was seriously injured. Contending that the platform was defective, Daniel and his wife Andrea filed this tort litigation under the diversity jurisdiction. After being served with process, EGA sent the papers to Meadowbrook Risk Management, which superintends EGA's insurance coverage. EGA anticipated that Meadowbrook would arrange for a defense, as the policy provides. North American Manufacturers Insurance Co. issued the policy; EGA deals with the insurer through Meadowbrook, which in the past had obtained counsel on behalf of both insurer and insured. After receiving notice, however, Meadowbrook did—nothing. It did not notify the insurer, did not engage counsel to represent EGA, and did not alert EGA to the need to protect its own interests; Meadowbrook just sat on its hands. Because no one filed an answer on EGA's behalf, the district court's clerk entered a default. Alerted to this by EGA, Meadowbrook's inactivity continued. When a magistrate judge recommended that plaintiffs receive $31.2 million in damages, EGA hired its own lawyer to see whether the situation could be salvaged.

Five months after its answer to the complaint had been due, EGA filed a motion to vacate the default, see Fed.R.Civ.P. 55(c), blaming Meadowbrook for the earlier inaction. The magistrate judge recommended that this motion be denied, observing that the best way to give Meadowbrook an incentive to take care is to amerce EGA, which could shift the expense to Meadowbrook. But the district judge concluded that the injury attributable to Meadowbrook's neglect is much less than $31 million, making the proposed award disproportionate to the wrong. In the judge's view, the lack of correspondence between EGA's limited fault and the $31 million award was "good cause" for reopening the case. The judge thus set aside the default, while holding open the possibility that a more appropriate sanction might be in order.

After discovery had been completed, a settlement was implemented via an offer of judgment under Fed.R.Civ.P. 68. The settlement provides that North American Manufacturers Insurance will pay the Simses the full amount remaining on EGA's policy, some $761,000. Plaintiffs reserve the right to appeal the question whether the default should have been set aside; the settlement provides that, if we reverse the district judge on this question and reinstate the default, then EGA would try to recoup any award in excess of $761,000 from Meadow-brook. (An order requiring the district court to reinstate the default would leave damages open, for the district judge did not rule on the magistrate judge's recommendation to award $31.2 million.) Finally, "in the event that the Court of Appeals declines to rule for any reason on all issues concerning [the default judgment], ... the parties are placed back into the same position they were in prior to [this settlement], i.e. awaiting trial in the district court".

Meadowbrook, having done nothing to protect its client EGA, decided to protect itself by intervening in the tort litigation. It opposed the settlement, arguing that the possibility of a $31 million award would cause it economic injury. One of its arguments was that Rule 68 had not been used properly. Rule 68 is designed to shift costs to a litigant who refuses a proper offer; here the offer was accepted, and how the parties got to "yes" is beside the point. But if Meadowbrook became a party, it could block the settlement by withholding its own assent. The district judge denied the motion to intervene, however, observing that Meadowbrook's liability, if any, could be addressed in a follow-on action if this court should reverse and a $31 million judgment ensue. Judgment was entered incorporating the terms of the Rule 68 offer.

Both the Simses and Meadowbrook have appealed—and Meadowbrook leads with the argument that we should dismiss the Simses' appeal for want of jurisdiction. If that happens, then the case will be tried on the merits in the district court. Meadowbrook apparently thinks that this would let it off the hook, but that's not so. If EGA were to prevail on the merits, or the Simses were to win less than $31 million, they could appeal from the final decision and ask that the default be reinstated and the magistrate judge's recommendation about damages be turned into a judgment. No matter. Jurisdiction is an appellate court's first order of business even if the person who had raised the question disserves his own interests, indeed even if no one has raised the question.

We have jurisdiction only if the judgment based on the settlement is a "final decision." 28 U.S.C. § 1291. This judgment leaves an issue to be resolved on appeal—but so does a conditional plea under Fed.R.Crim.P. 11(a)(2) that reserves an issue for appellate decision. Courts regularly entertain disputes from such dispositions without doubting their finality. A settlement that reserves an issue for appeal is final not only because it completely resolves the litigation if no appeal ensues, but also because an affirmance leaves that disposition in place. No further litigation will occur. The civil rules lack a provision comparable to Rule 11(a)(2) of the criminal rules, but this does not affect jurisdiction. So we held in Downey v. State Farm Fire & Casualty Co., 266 F.3d 675, 682-83 (7th Cir.2001), when taking jurisdiction of an appeal from a judgment based on a conditional settlement in a civil case.

Does the settlement's provision that the litigation resumes if we refuse to take jurisdiction alter matters? Not at all. We read this provision to say no more than that, if we think the decision not final, then it must be not final and the litigation must continue. The same proviso is implied in a conditional plea under Rule 11(a)(2); if resolving the reserved issue on appeal is impossible, then the plea must be set aside and the prosecution continue. What is implied under Rule 11(a)(2) has been made explicit in this judgment.

Meadowbrook invokes a line of decisions exemplified by Horwitz v. Alloy Automotive Co., 957 F.2d 1431 (7th Cir.1992). Many civil litigants would like to have interlocutory appellate resolution of some important issue. But when the issue does not meet the standard for certification under 28 U.S.C. § 1292(b), and it is not feasible to enter judgment with respect to a separate claim or party under Fed.R.Civ.P. 54(b), it is not possible to obtain an appellate decision until the case is over. Crafty litigants thought that they could evade the limits on interlocutory review by asking the district judge to dismiss the suit with leave to reinstate once an appeal had been resolved. Then they argued that nothing remained pending in the district court to spoil the "finality" of the decision. Horwitz and its successors refuse to go along with that ploy, because the rest of the litigation remains in the background, ready to resume as soon as the appeal ends. In evaluating "finality," we held in Horwitz, it is essential to look at the whole picture, including claims that have been put on the back burner through a dismissal-with-leave-to-reinstate procedure.

Nothing in this case has been dismissed with leave to reinstate, however. Nobody is trying to pull a fast one. If we affirm, the case is over and the Simses receive $761,000. If we reverse, the default is reinstated (avoiding any dispute about liability) and the Simses will be entitled to an award of damages that may be substantially higher. Only if we refuse to decide does the case go to trial on the merits. That's just what should happen, because the only way we can refuse to act is if the judgment isn't final, and then the case must still be ongoing. The gimmick that prevented an appeal in Horwitz—that the litigation would continue no matter what happened on appeal—has not been reused. So the decision is final and appealable. We shall decide the question presented, and proceedings will be over.

A default judgment is a sanction for misconduct during the litigation. Appellate review of decisions to impose, or withhold, sanctions is deferential. See, e.g., Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 399-400, 110 S. Ct. 2447, 110, L.Ed.2d 359 (1990); National Hockey League v. Metropolitan Hockey Club, Inc., 427 U.S. 639, 93 S. Ct. 2778, 49 L. Ed. 2d 747 (1976); Pretzel & Stouffer v. Imperial Adjusters, Inc., 28 F.3d 42, 45 (7th Cir.1994). That standard of review pretty much resolves this appeal, for it would be unimaginable to label as an abuse of discretion a district judge's decision that some sanction less than $31 million is appropriate for a wrong that apparently caused no prejudice to the adverse party. See Mommaerts v. Hartford Life & Accident Insurance Co., 472 F.3d 967, 2007 WL 38406 (7th Cir. Jan. 8, 2007). Cf. United States v. McLaughlin, 470 F.3d 698 (7th Cir.2006).

Defaults may be set aside for "good cause". Damages disproportionate to the wrong afford good cause for judicial action, even though there is no good excuse for the defendant's inattention to the case. Rule 55(c) requires "good cause" for the judicial action, not "good cause" for the defendant's error; as used in this Rule, the phrase is not a synonym for "excusable neglect." See Redfield v. Continental Casualty Corp., 818 F.2d 596, 601 (7th Cir. 1987). (Another way to see this is that Rule 55(c) uses the "good cause" standard for relief before judgment has been entered, while referring to the standard under Rule 60(b) for relief after judgment. Rule 60(b) allows relief on account of mistake and inadvertence in addition to excusable neglect; the "good cause" standard in Rule 55(c) must be easier to satisfy.)

In Degen v. United States, 517 U.S. 820, 116 S. Ct. 1777, 135 L. Ed. 2d 102 (1996), the Supreme Court held it abuse of discretion to default a litigant in a $5.5 million civil suit for wilful failure to appear in court, given the availability of lesser sanctions. (The litigant was a fugitive in a criminal prosecution.) Likewise a $31 million sanction would be excessive in this suit: EGA's misconduct was negligent (reckless at worst) rather than deliberate, and the injury (if any) to its adversary was negligible. Like damages in civil litigation, the sanction should fit the offense. That's a major reason why district judges should not enter defaults precipitately. See, e.g., Bleitner v. Welborn, 15 F.3d 652 (7th Cir.1994); Philips Medical Systems International B.V. v. Bruetman, 8 F.3d 600 (7th Cir. 1993); cf. Ball v. Chicago, 2 F.3d 752 (7th Cir.1993).

Our point is not that EGA should be let off easy because Meadowbrook is principally responsible. A litigant bears the consequences of errors by its chosen agent. See, e.g., Pioneer Investment Services Co. v. Brunswick Associates L.P., 507 U.S. 380, 396-97, 113 S. Ct. 1489, 123 L. Ed. 2d 74 (1993); United States v. 7108 West Grand Avenue, 15 F.3d 632 (7th Cir. 1994). Cf. United States v. Boyle, 469 U.S. 241, 105 S. Ct. 687, 83 L. Ed. 2d 622 (1985). Penalizing the litigant, who can shift costs to the agent in turn, is the best way to ensure that the agent takes adequate care. So we proceed as if all fault is imputed to EGA. Nor do we imply that defendants in multi-million-dollar cases may treat the proceedings as nuisance litigation; a firm in EGA's position should invest more in the oversight of its insurer when the plaintiff demands $10 million than when the ad damnum is $100,000.

Still, delay that imposes slight injury does not call for multi-million-dollar awards. EGA's errors in managing its defense of this litigation did not cause Daniel Sims to fall on his head; that is the source of the $31 million loss (if the magistrate judge estimated the damages correctly). What EGA did (better, neglected to do) extended this suit by a few months and perhaps caused the Simses some anxiety and marginal legal expenses. A court can compensate aggrieved litigants for such losses directly; the district judge here did not abuse his discretion in concluding that entry of default would be overkill.

Note that we have referred to the district judge's discretion, not the magistrate judge's. Even if we suppose, as the Simses contend, that the magistrate judge could have resolved this issue under Fed. R.Civ.P. 72(a)—which seems unlikely, for default concludes the merits, while Rule 72(a) covers only "nondispositive matters"—a district court is not obliged to give magistrate judges the maximum authority such a non-Article-III officer may wield. The litigants did not consent to final decision by a magistrate judge, see 28 U.S.C. § 636(c), so the district judge remained in charge and was entitled to make an independent decision, which he did.

Because the district judge's decision to set aside the default was not an abuse of discretion, the Simses will receive $761,000, and Meadowbrook is not at risk of an excess judgment. There is no remaining litigation into which it could intervene. For what it is worth, however, we think that the district judge acted sensibly in denying the motion to intervene. Third-party liability is best handled in third-party actions. Allowing Meadowbrook to intervene would make no more sense than allowing lawyers to intervene as parties whenever some question about the competence of their work is raised. Just as those questions should be resolved in separate malpractice actions, disputes about insurers' (or insurance agents') liability belong in separate litigation.

AFFIRMED.

CUDAHY, Circuit Judge, concurring.

I agree completely with the majority, but write separately to comment in greater detail on what I think is an important aspect of the confusing issue of appellate jurisdiction. The consent judgment here specified that in the event that we "decline to rule for any reason on all of the issues concerning" the judgment (presumably because we hold that the order is nonfinal and dismiss the appeal),1 then all of the parties' claims and defenses "reignite": the order "will be considered void and unenforceable ... and the parties are placed back into the same position they were in prior to entering this offer of judgment." In this way, the judgment permits the parties to test without risk whether the order is an appealable "final decision[ ]," 28 U.S.C. § 1291, that is, without committing themselves to any consequences if we determine that it is not.

Superficially, such a disposition might seem similar to the nonfinal orders discussed in the majority opinion, which are characterized by parties attempting to test parts of their case on appeal without risking consequences involving other parts of their case. See Majority Op. at 867-68, citing Horwitz v. Alloy Automotive Co., 957 F.2d 1431 (7th Cir.1992). Orders are not final unless they leave nothing for the district court to do (absent whatever is called for by reversal on appeal or by the application of Rule 59(e) or 60). Green Tree Fin. Corp. v. Randolph, 531 U.S. 79, 86, 121 S. Ct. 513, 148 L. Ed. 2d 373 (2000). Consequently, a consent order is not final unless the parties "gamble" by "stak[ing] the[ir] entire case" on the outcome of the appeal from it. First Health Group Corp. v. BCE Emergis Corp., 269 F.3d 800, 801-02 (7th Cir.2001). Attempting to provide for more litigation in the event of reversal than the holding on appeal calls for renders the order nonfinal. See Union Oil Co. of Cal. v. John Brown E & C (Unocal), 121 F.3d 305 (7th Cir.1997).

As the majority points out, what distinguishes from cases such as Horwitz the parties' present attempt to risklessly test appellate jurisdiction is that whether an order purporting to resolve a case is final is not itself part of the case; obviously appellate jurisdiction is not an issue before the district court. The present case involves the Sims' product liability claims against EGA, and the consent judgment definitively resolved them. The outcome will be altered and the litigation will reignite in the event that we find we lack jurisdiction. While modification of a final order is strictly circumscribed, see Fed.R.Civ.P. 59(e) & 60, nonfinal orders are generally modifiable, see, e.g., Fed.R.Civ.P. 54(b) ("[T]he order or other form of decision is subject to revision at any time before the entry of judgment adjudicating all the claims and the rights and liabilities of all the parties."). If the reignition provision were to be activated, it would be perfectly supported by our holding of finality, rather than at war with it. See Majority Op. at ____ ("[I]f we think the decision not final, it must be not final and the litigation must continue.").

By limiting appealable orders, 28 U.S.C. § 1291 serves, among other things, to promote efficient judicial administration and conserve scarce judicial resources. Cunningham v. Hamilton County, 527 U.S. 198, 204, 119 S. Ct. 1915, 144 L. Ed. 2d 184 (1999); Cobbledick v. United States, 309 U.S. 323, 325, 60 S. Ct. 540, 84 L. Ed. 783 (1940); Unocal, 121 F.3d at 310. But "efficient judicial administration" is not synonymous with "forbidding as many appeals as possible." Permitting parties to risklessly test appellate jurisdiction may encourage some improper appeals from nonfinal judgments, but in other cases, such as this one (where an order is clearly appealable absent the provision allowing for reignition), allowing the test will hasten final resolution.

At any rate, Congress has made its determination as to which appeals will best promote efficient adjudication; it has prescribed appeals from "final decisions." Decisions are final, appropriately enough, if they end litigation in the district court, leaving it nothing more to do. Appellate jurisdiction is not an issue before the district court, and the provisions of the judgment triggered by lack of appellate jurisdiction give the district court no task to perform. The consent judgment here is therefore final and appealable.