DIANE P. WOOD, Circuit Judge. This case began almost three decades ago when inmates at what is now the Pendleton Correctional Facility in Indiana filed a class action against the state claiming that certain conditions at the prison violated their constitutional rights. They obtained some of the relief they sought in an injunction that was affirmed by this court in French v. Owens, 777 F.2d 1250, 1258 (7th Cir. 1985), and the prison has operated under that injunction, as modified from time to time, ever since. The present action arose when the State of Indiana decided to take advantage of the 1996 Prison Litigation Reform Act ("PLRA") and petition to terminate the injunction. The merits of that effort, however, are not before us at this time. Instead, we must decide whether the so-called "automatic stay" provision of the PLRA, codified at 18 U.S.C. §3626(e)(2), applies, and if it does, whether it is constitutional.

In order to place this case in context, we begin with a brief description of the PLRA as it affects injunctions addressing prison conditions. The part of the statute with which we are concerned addresses the subject of "[a]ppropriate remedies with respect to prison conditions." 18 U.S.C. §3626. Subpart (a)(1) provides that a federal court must limit prospective relief with respect to prison conditions in a variety of ways:

The court shall not grant or approve any prospective relief unless the court finds that such relief is narrowly drawn, extends no further than necessary to correct the violation of the Federal right, and is the least intrusive means necessary to correct the violation of the Federal right.

18 U.S.C. §3626(a)(1).

Recognizing that many institutions are already operating under existing injunctions, the statute also provides a way for the prison authorities to bring their federally imposed obligations into line with the limitations of §3626(a)(1). Part (b) sets out a road map for the termination of prospective relief that has either outlived its usefulness, or that violates the (a)(1) conditions. Under the authority of §3626(b)(1), prospective relief is terminable upon motion of any party or intervener within two years after the court granted the relief (or two years after the date of enactment of the PLRA), or one year after the court denied a request to terminate relief. Subpart (b)(2), which has come to be known as the "immediate termination" provision of the statute, establishes the defendant's or intervener's right to relief. Because this is the basis for the state's petition, we set it out in its entirety:

(2) Immediate termination of prospective relief.- -In any civil action with respect to prison conditions, a defendant or intervener shall be entitled to the immediate termination of any prospective relief if the relief was approved or granted in the absence of a finding by the court that the relief is narrowly drawn, extends no further than necessary to correct the violation of the Federal right, and is the least intrusive means necessary to correct the violation of the Federal right.

18 U.S.C. §3626(b)(2). Hard on the heels of this provision is another that sets forth an exception to the entitlement to "immediate" termination:

(3) Limitation.- -Prospective relief shall not terminate if the court makes written findings based on the record that prospective relief remains necessary to correct a current and ongoing violation of the Federal right, extends no further than necessary to correct the violation of the Federal right, and that the prospective relief is narrowly drawn and the least intrusive means to correct the violation.

18 U.S.C. §3626(b)(3). Part (c) of the PLRA addresses settlements, basically saying that a court may not enter a settlement in the form of a consent decree unless the settlement conforms to the statutory limitations, but that the parties are free to conclude any private settlement agreement they wish, as long as that agreement is not directly enforceable by the court (other than by reinstatement of the case). Part (d) makes clear that the PLRA's limitations do not apply to relief entered by a state court based solely upon claims arising under state law.

Finally (for our purposes) is the automatic stay provision, part (e). It begins innocuously enough in subpart (e)(1), by calling for the court to rule promptly on any motion to modify or terminate prospective relief. The problems arise with subpart (e)(2), which provides as follows:

(2) Automatic stay.- -Any motion to modify or terminate prospective relief made under subsection (b) shall operate as a stay during the period- -

(A)(i) beginning on the 30th day after such motion is filed, in the case of a motion made under paragraph (1) or (2) of subsection (b); or

(ii) beginning on the 180th day after such motion is filed in the case of a motion made under any other law; and

(B) ending on the date the court enters a final order ruling on the motion.

18 U.S.C. §3626(e)(2). Since 1997, it has been possible for the court to order a modest extension of time before the automatic stay goes into effect, if the court takes advantage of subpart (e)(3):

(3) Postponement of automatic stay.- -The court may postpone the effective date of an automatic stay specified in subsection (e)(2)(A) for not more than 60 days for good cause. No postponement shall be permissible because of general congestion of the court's calendar.

18 U.S.C. §3626(e)(3). See Pub. L. No. 105-119 §123, 11 Stat. 2440, 2470 (1997) (adding this language). Finally, under subpart (e)(4) the statute expressly provides that an order "staying, suspending, delaying, or barring the operation of the automatic stay described in paragraph (2)" other than the orders authorized by (e)(3) may be appealed pursuant to 28 U.S.C. §1292(a)(1).

With that background in mind, we now turn to what happened in this case. On June 5, 1997, Warden Jack R. Duckworth of the Pendleton Correctional Facility, along with Indiana officials Bruce Lemmon and Edward L. Conn (to whom we refer collectively as the state), filed a Motion To Terminate Decree, relying on §3626(b)(1)(A) and (b)(2). In response, on June 30, 1997, the prisoner class ("the prisoners") filed a Motion for a Temporary Restraining Order or Preliminary Injunction, in which they asked the court to stay the automatic stay provision of §3626(e)(2), which they described in paragraph 2 of their motion. They also filed a memorandum in support of the motion, in which they addressed the four standards that normally govern the issuance of a preliminary injunction: (a) likelihood of success on their argument that the automatic stay provision of §3626(e)(2) would be found unconstitutional; (b) irreparable harm to the prisoners if the "automatic termination" took effect; (c) lack of harm to the defendants if the "automatic termination" was stayed; and (d) the public interest. The prisoners' memorandum concluded with a request that "the Court should enter a temporary restraining order or a preliminary injunction staying the operation of the automatic termination provision of the PLRA." On the same day, the prisoners also filed a separate Response to Defendants' Motion to Terminate Decree, in which they set forth their position that the termination provisions of §§3626(b)(2) and (b)(3) are unconstitutional.

On July 3, 1997, the district court granted the temporary restraining order the prisoners had requested and scheduled a hearing on July 10 for fuller consideration of the preliminary injunction motion. On July 11, following the hearing, the district court entered an order converting the TRO into a preliminary injunction. The state has now appealed from that July 11 order.

Our review of the district court's actions is complicated by the fact that the court's orders granting the TRO and preliminary injunction can be read as confusing the §3626(e)(2) automatic stay provision with the §3626(b)(2) immediate termination provision. (This confusion seems to have originated with the parties themselves. We note that although the prisoners sought to stay the "automatic termination" provision, there is no such thing in the PLRA. There is one section providing for immediate termination, and another requiring an automatic stay.) Even though the prisoners had not asked for a TRO against the (b)(2) immediate termination provision, the July 3 order recites that it "enjoin[s] and prohibit[s] the automatic termination [sic] provision of the Prison Litigation Reform Act from taking effect." Further, the July 11 order explains that the court was converting the TRO into a preliminary injunction "for the principal reason that the Court believes that 18 U.S.C. §3626(b)(2) is clearly unconstitutional as found by other courts." The July 11 order concludes that "[a]ccordingly, there shall be no stay of prospective relief in this matter."

Although we considered remanding this case to the district court to find out what it really meant, in the end we concluded that such a step was unnecessary. A review of the entire record (including the transcript of the July 10 hearing) convinces us that the district court intended to enjoin the (e)(2) automatic stay provision, not the (b)(2) immediate termination provision. True, the orders referred to the "automatic termination" provision and to 28 § U.S.C. 3626(b)(2). Several points convince us, however, that the district court simply cited the wrong statutory section, a mistake to which we need not attach any significance unless it affects the substantial rights of the parties, see Fed. R. Civ. P. 61. First, the motion before the court and the arguments presented at the hearing on the preliminary injunction addressed only the automatic stay of (e)(2), not the immediate termination required by (b)(2). There was a fair amount of discussion of (b)(2) at the hearing, but this was in the context of the likelihood that the prisoners could defeat the state's underlying petition for immediate termination. Second, the record shows that proceedings have continued in the district court on the question whether the decree should be terminated pursuant to §3626(b)(2), suggesting that the lower court is of the view that it did not finally resolve the termination issue. Finally, and most important, the last paragraph of the court's injunction sets forth in plain English what it was doing. Stripped of citations, the only thing that paragraph does is to refuse to permit a stay of prospective relief to go into effect. In substance, this means the injunction addresses the (e)(2) problem, not the (b)(2) problem.

Even though the question of the constitutionality of (b)(2) is not before this panel, it is presented in Berwanger v. Cottey, No. 98-3107, which this court has also decided today. In Berwanger, we join the overwhelming majority of our sister circuits and hold that §3626(b)(2) may be applied to existing consent decrees. See, e.g., Imprisoned Citizens Union v. Ridge, 1999 WL 93802 (3d Cir. Feb. 25, 1999); Tyler v. Murphy, 135 F.3d 594 (8th Cir. 1998); Hadix v. Johnson, 133 F.3d 940 (6th Cir. 1998); Dougan v. Singletary, 129 F.3d 1424 (11th Cir. 1997); Inmates of Suffolk County Jail v. Rouse, 129 F.3d 649 (1st Cir. 1997); Benjamin v. Jacobson, 1999 WL 188240 (2d Cir., March 23, 1997) (en banc); Gavin v. Branstad, 122 F.3d 1081 (8th Cir. 1997); Plyler v. Moore, 100 F.3d 365 (4th Cir. 1996). Cf. Taylor v. United States, 143 F.3d 1178 (9th Cir. 1998) (finding §3626(b)(2) unconstitutional), opinion withdrawn and en banc rehearing granted, 1998 WL 775245 (9th Cir., Nov. 3, 1998).

That leaves the question whether the district court was empowered to enjoin the operation of §3626(e)(2)'s automatic stay, and if it was not, whether that subsection is constitutional. Bearing in mind that we should try to avoid constitutional questions if we can, we address the question of statutory interpretation first.

The only other court of appeals to have considered the scope and constitutionality of (e)(2) is the Sixth Circuit, in its opinion in Hadix v. Johnson, 144 F.3d 925 (6th Cir. 1998). That court concluded that if (e)(2) were read to mean what it apparently says- -that is, as a legislative command that a stay of prospective relief occurs as a matter of law no later than 90 days after the state's petition is filed, and that this legislative stay continues in effect until the findings required by (b)(3) have been made- -then it would be "an unconstitutional incursion by Congress into the powers reserved for the Judiciary." Id. at 937. In order to avoid a finding of unconstitutionality, the court chose to adopt the position that the Justice Department had urged and concluded that the courts "retain the power to suspend the automatic stay in accordance with general equitable principles." Id. By that, the Sixth Circuit meant that the automatic stay of (e)(2) could itself be stayed by a court using the traditional standards for issuing a preliminary injunction. Id. at 945.

While the decisions of the district courts do not create binding precedents (particularly not those within the Sixth Circuit, which are now bound by Hadix), it is nevertheless noteworthy that a number of judges have agreed with the Sixth Circuit that subpart (e)(2) is unconstitutional if it is a legislative stay, but they have not agreed that the statute can be saved by a narrowing interpretation. See United States v. Michigan, 989 F.Supp. 853 (W.D. Mich. 1996); Glover v. Johnson, 957 F. Supp. 110 (E.D. Mich. 1997); Hadix v. Johnson, 933 F.Supp. 1362 (W.D. Mich. 1996) (Enslen, C.J.); Hadix v. Johnson, 933 F.Supp. 1360 (E.D. Mich. 1996) (Feikens, J.). The question is one of first impression in this circuit, and we have benefitted from the careful consideration that our colleagues on the federal bench have devoted to it.

We are well aware of the rule requiring courts to construe statutes consistently with the Constitution, if the language will bear any such construction. Edward J. DeBartolo Corp. v. Florida Gulf Coast Bldg. & Constr. Trades Council, 485 U.S. 568, 575 (1988), citing Hooper v. California, 155 U.S. 648, 657 (1895); N.L.R.B. v. Catholic Bishop of Chicago, 440 U.S. 490, 500 (1979). But the qualification that the language must be able to bear the constitutional interpretation is an important one. Courts cannot redraft statutes so that they read the way Congress might have written them, or should have written them. Instead, we must taken the laws as they are given to us and work with them.

As was the case before the Sixth Circuit, three different views of (e)(2) have been urged on us: the prisoners argue that it is unconstitutional on several grounds; the state defends its constitutionality as a mere regulation of procedure; and the United States is of the view that (e)(2) would be an unconstitutional violation of either the separation of powers doctrine or the prohibition against legislative suspension of a particular judgment if (e)(2) really required automatic stays. The United States argues, however, that the automatic stay provision should be read to instruct a court reviewing prospective relief that the court should- -but is not required to- -stay the relief, an interpretation that avoids the constitutional flaws.

Before we delve too far into these points, we must decide which reading of subpart (e)(2) is correct: that of the prisoners and the state, on the one hand, or that of the United States, on the other. If we agree with the Sixth Circuit that the (e)(2) stay is discretionary in the final analysis, then there would be no need to explore the constitutional issues further. The statute would be reduced to something that placed the burden of moving for a stay or preliminary injunction on the prisoners rather than the state, but once the issue was before the court the normal equitable considerations would determine whether existing decree provisions remained in place pending a decision on a termination petition, or if they should be modified.¹ Subpart (e)(4) of the statute would in addition make it clear that appellate review under 28 U.S.C. §1292(a)(1) was available no matter what the court called its order "staying, suspending, delaying, or barring the operation of the automatic stay described in paragraph (2). . . ."

Although we have the highest regard for our Sixth Circuit colleagues and the concerns that motivated them to adopt the Justice Department's view of (e)(2), we cannot agree that the language of that subpart can be pushed this far. First, Congress used unequivocal words when it drafted (e)(2). A motion to modify or terminate prospective relief made under part (b) shall operate as a stay. Congress specified that the stay would be automatic. Finally, it specified not only a clear starting point, but also the ending point for the stay. Even though we do not lightly assume that Congress meant to restrict the equitable powers of the federal courts, we find it impossible to read this language as doing anything less than that.

Recognizing that it was hard to find internal support for its reading within (e)(2) alone, the Sixth Circuit relied in part on the fact that (e)(4) makes orders staying or otherwise barring the automatic stay automatically appealable. Hadix, 144 F.3d at 936. Why would Congress have included this in the statute, they reasoned, if it did not anticipate that courts would continue to have the power to issue equitable relief against the stay? In our view, there is an answer to that question. The drafters of the PLRA realized that they were skating close to the line in (e)(2), and they wanted to ensure that the issue that is now before us could be resolved in an interlocutory appeal. The fact that a district court's effort to stay the (e)(2) stay can be appealed says nothing about what an appellate court must do once it has the case. Congress undoubtedly hoped that the appellate courts would reject these district court efforts and enforce the regime that (e)(2) sets up. That, however, is possible only if (e)(2) as we believe it must be read is constitutional, and so we now turn to that question.

The state (which as we said agrees with our reading of the statute) has defended (e)(2) as nothing more than the establishment of a procedure that assures prompt review of prospective injunctive relief in prison litigation. All the prisoners need do, the state asserts, is to come forward with evidence before the 30th day (or the 90th day, if the court grants an extension of time) showing that conditions at the prison require continuation of some or all of the prospective injunctive relief. In other words, the prisoners must make a complete showing on the merits of the termination petition, the state must complete its rebuttal of that showing, and the court must make its ruling, all within the 90-day period. The state analogizes (e)(2) to the 10-day limit on TROs found in Fed.R.Civ.P. 65(b), to the five-day limit on TROs under the Norris-LaGuardia Act, 29 U.S.C. §107, and to the automatic stay in bankruptcy cases, 11 U.S.C. §362.

There are important differences, however, between the automatic stay of (e)(2) and the statutes on which the state relies. Both the 10-day limit on TROs found in Rule 65 (b) and the analogous limit in the Norris-LaGuardia Act respond to the particular problems of ex parte proceedings. Nothing in those laws purports to restrict the power of the district court to enter a preliminary injunction that preserves the status quo beyond the period of time allowed. The automatic stay in bankruptcy also does not help the state's case- -indeed, if anything it undermines it. This is because the automatic stay of §362, which is triggered by the filing of a petition, in bankruptcy, see 11 U.S.C. §362(a), has the effect of preserving the court's equitable powers over the entirety of the bankruptcy estate, not superseding or undermining them. In essence, it freezes everything external to the bankruptcy proceeding, including private transactions and state court litigation, and it has the effect of requiring other federal court proceedings to take a back seat to the federal bankruptcy court. Furthermore, litigants routinely petition the bankruptcy court to modify the stay, and there is no hint that Congress did not want the court to exercise that power (as there is with the PLRA). Allocations of power within the federal court system are a common feature of procedural legislation, and principles of federal supremacy account for the effect that the automatic stay has on state court proceedings.

The automatic stay of (e)(2), in contrast, operates directly on the internal adjudication of a case in federal court. It strips from the court the authority to decide whether the status quo (defined by the earlier decree the court entered that required prospective relief) should be continued or modified pending the court's decision on the immediate termination petition. It does so in a way that leaves the power to continue the decree entirely in the hands of the party that files the motion for termination. This is so because the automatic stay must take effect no later than the 90th day after the petition is filed unless the court has issued a final order on the termination motion. Yet the state need only drag its feet or confront genuine difficulty in responding to requests for information that is relevant to the question whether the decree continues to be necessary, as defined by (b)(2) and (b)(3), in order to win its stay. Given the complexity of much prison litigation, we would be reluctant to try to address the problem of delay through satellite litigation over whether the state was "really" acting in bad faith or not, and imposing adverse fact-findings on a state that was trying to manipulate the process so that it won an automatic stay. We are also concerned that such an approach would once again read too much into the statute. Section 3626 constrains the authority of the district courts to impose and sustain prospective relief. It says nothing about conscripting states into this process, and we see nothing in the overall tenor of §3626 that would justify superimposing on it a rule that a state risks the denial of its motion for termination on the merits if it does not (or cannot) comply with the (e)(2) time limits.

We do not suggest that Congress cannot prescribe rules of practice and procedure for the federal courts, and, contrary to the dissent's dire predictions, nothing in this opinion in any way threatens the ordinary time limits that pervade both federal procedural rules and statutes. First, it is plain that Congress has the power to prescribe procedural rules. See Hanna v. Plummer, 380 U.S. 460, 472 (1965); Sibbach v. Wilson & Co., 312 U.S. 1, 9 (1941). Many time limits are subject to adjustment, and thus are irrelevant for present purposes. Some both define the jurisdiction of a court and establish when a judgment is final, such as the time for filing an appeal from a final judgment, Fed.R.App.P. 4(a), (b), or the time within which a motion to correct a sentence must be filed, Fed.R.Crim.P. 35(c). But that is not what (e)(2) does, and it is important to appreciate just how unusual a provision it is. As the Sixth Circuit also recognized, temporal strictures upon substantive judicial decision making are scarce. Hadix, 144 F.3d at 943-44 n.15. Where such rules and statutes tie a judicial outcome to a time restriction, the court usually retains discretion to override the restriction for good cause. Id. For example, the Speedy Trial Act sets forth time limits for criminal trials, including a requirement that if the Act's deadlines are violated, the charges must be dismissed. See 18 U.S.C. §3161; see also United States v. Brainer, 691 F.2d 691, 695-99 (4th Cir. 1982) (upholding the Speedy Trial Act against a challenge that the Act violates the doctrine of separation of powers). Unlike §3626(e)(2), however, the Act contains a long list of exceptions, see 18 U.S.C. §3161(h), including a broad provision that authorizes judges to exclude from the time calculations "any period of delay" upon "finding[] that the ends of justice served by taking such action outweigh the best interest of the public and the defendant in a speedy trial." 18 U.S.C. §3161(h)(8)(A). See also Fed.R.Civ.P. 60(b) (authorizing courts "upon such terms as are just" to relieve a party from a final judgment, including a judgment for prospective relief).

Similarly, the Antiterrorism and Effective Death Penalty Act ("AEDPA") states that "[t]he Court of Appeals shall grant or deny the authorization to file a second or successive application [for a writ of habeas corpus] not later than 30 days after the filing of the motion." 28 U.S.C. §2244(b)(3)(D). For many of the same reasons that animate us today, however, the courts of appeals have ruled that this time limit may be modified if the court finds it necessary. See, e.g., In re Siggers, 132 F.3d 333, 336 (6th Cir. 1997) (reading the language of §2244(b)(3) as "hortatory or advisory rather than mandatory," in order to avoid constitutional difficulties); Galtieri v. United States, 128 F.3d 33, 36-37 (2d Cir. 1997) (ruling that §2244(b)(3) must be applied "flexibly," and concluding that the courts should not forego "reasoned adjudication" in the small number of cases that cannot be resolved within 30 days): In re Vial, 115 F.3d 1192, 1194 n.3 (4th Cir. 1997) (noting that the court exceeded the 30-day limit but concluding that the importance of the issue justified the delay). As we noted earlier, this is precisely the kind of flexibility that we believe is foreclosed to the courts under the language of §3626(e)(2).²

The fact that Congress can impose time limits on executive agencies is of little assistance here. Unlike temporal limitations on judicial decision making, such constraints on agency action are prevalent throughout administrative law. See, e.g., 7 U.S.C. §2a(iv)(II) (setting s 30-day time limit for Security and Exchange Commission's review of a board of trade's application for designation as a contract market). Where agencies are involved, the judgments subject to legislative encroachment were not rendered by Article III courts but by entities that the Constitution places under the control of Congress. The distinction between those two situations motivated the Court's ruling in Paramino Lumber Co. v. Marshall, 309 U.S. 370 (1940), cited with approval in Plaut v. Spendthrift Farm, Inc., 514 U.S. 211, 232 (1995), where it upheld a private bill that reopened an administrative order, noting that the order was not an adjudication and therefore the legislative act did not constitute "an excursion of Congress into the judicial function" or "affect[] judicial judgments." 309 U.S. at 381 & n.15. Last, we think there is an important difference between legislation that affects prospective relief, and legislation that prevents the court from preserving the status quo in whole or in part during the pendency of a suit. The former is permissible, as Plaut observed, but the latter touches upon the heart of the adjudicative process and as such is reserved for the Judicial Branch of government- -a point on which we now elaborate.

The prisoners have urged that (e)(2) violates Article III and the separation of powers principle, as well as the due process clause, and thus that it must be declared unconstitutional. Such a finding would not have any effect on the remainder of the PLRA, because Congress included an express severability clause in the statute. See Pub. L. No. 104-134, Title I, §101(a), Apr. 26, 1996, 110 Stat. 1321-77, renumbered Title I, Pub. L. No. 104-140, §1(a), May 2, 1996, 110 Stat. 1327. We conclude, as did our colleagues in the Sixth Circuit when they analyzed the statute according to the reading we have adopted, that (e)(2) violates the separation of powers principle because it is a direct legislative suspension of a court order. See Hadix, 144 F.3d at 939. Because we believe that (e)(2) represents an unconstitutional legislative encroachment into the powers reserved to the judiciary, we do not need to reach the prisoners' due process arguments.

Under the Supreme Court's decision in Plaut, we know that Congress cannot vest review of the decisions of Article III courts in officials of the executive branch. See 514 U.S. at 218, citing Hayburn's Case, 2 U.S. (2 Dall.) 409 (1792), and Chicago & Southern Air Lines, Inc. v. Waterman S.S. Corp., 333 U.S. 103 (1948). The Sixth Circuit found, and we agree, that there is no principle under which the legislative branch should enjoy a privilege of reviewing particular decisions of Article III courts that the executive branch does not have. See Hadix, 144 F.3d at 940. As the Supreme Court explained in Plaut, "the Framers crafted [Article III] . . . with an expressed understanding that it gives the Federal Judiciary the power, not merely to rule on cases, but to decide them, subject to review only by superior courts in the Article III hierarchy." 514 U.S. at 218-19. Yet (e)(2) places the power to review judicial decisions outside of the judiciary: it is a self-executing legislative determination that a specific decree of a federal court- -here the decree addressing conditions at Pendleton- -must be set aside at least for a period of time, no matter what the equities, no matter what the urgency of keeping it in place. This amounts to an unconstitutional intrusion on the power of the courts to adjudicate cases.

Unlike the Sixth Circuit, we also find that (e)(2) violates the principle articulated in United States v. Klein, 80 U.S. (13 Wall.) 128 (1871). In Klein, Congress passed a statute providing that individuals whose property was seized during the Civil War could recover the property by showing that they had not offered "aid or comfort" to the enemy during the war. Klein, 80 U.S. at 131. The Supreme Court subsequently held that a presidential pardon was a sufficient showing that an individual was not a Confederate sympathizer. Reacting to this decision, Congress adopted legislation that required the court to consider a presidential pardon as conclusive evidence of the person's disloyalty to the United States and to dismiss appeals in cases seeking to recover the property of such persons. The Klein Court struck down this law, holding that Congress does not have the power to impose a rule of decision for pending judicial cases, apart from its power to change the underlying applicable law.

Applying Klein to 18 U.S.C. §3626(e)(2), the Sixth Circuit focused on the broader remedial provisions of the PLRA and found that the automatic stay did not mandate a rule of decision. Hadix, 144 F.3d at 940. The PLRA simply confined the relief that may be ordered in a prison conditions case to measures strictly designed to address violations of federal law. Id. But, as we have been emphasizing, (e)(2) does not directly implicate the final decision on the merits of the (b)(2) termination motion. Instead, it addresses what should happen during the pendency of the case. For that time period, the statute does mandate a particular rule of decision: the prospective relief must be terminated. In our view, this falls comfortably within the rule of Klein, and as such, it exceeds the power of the legislative branch.

Our determination that the automatic stay provision is not enforceable does not mean that courts should not try to conform their conduct to it. Indeed, the contrary is true: we hold that the district courts must conform their actions to the time limits in §3626(e)(2) unless compelling reasons for setting them aside can be articulated. On an interlocutory appeal under §3626(e)(4), this court has the power to require either prompt action or a stay of prospective relief by finding that a recalcitrant judge has abused her discretion. In passing the PLRA, Congress sought to quell the perceived tendency of federal courts to micro-manage state prisons by limiting the availability of consent decrees and other judicially-imposed prospective relief. See H.R. Rep. No. 21, 104th Cong., 1st Sess. 9 (1995) (noting that federal courts have "used these consent decrees to intrude into a state criminal justice system and seriously undermine the ability of the local justice system to dispense any true justice"). The PLRA accomplishes this goal, in part, by "includ[ing] provisions that will guard against court-ordered [remedies] dragging on and on, with nothing but the whims of federal judges sustaining them." H.R. Rep. No. 21, at 8.

It may be, however, that in some cases the courts will not be able to carry out their adjudicative function in a responsible way within the time limits imposed by (e)(2). See Hadix, 144 F.3d at 944. Given the command of the PLRA to tailor relief to the least restrictive alternative, and to take every step to ensure that an injunction does not stray beyond the requirements of federal law, the district courts will have a complex task on their hands. Some decrees under review will have been the result of years of litigation, and in considering whether termination is proper under §3626(b)(2), or whether newly tailored relief should continue under §3626(b)(3), the court may need not only to review a massive record, but also to take new evidence. In many other cases, like this one, the decree will be the result of a settlement between the parties and the record may be far too scant to make the required determinations. See generally Rufo v. Inmates of Suffolk County Jail, 502 U.S. 367, 378 (1992) (holding that a Fed.R.Civ.P. 60(b) motion may be used to seek modification of a consent decree). The court will have the obligation to supplement the record so that its final orders comply with the statute, and both sides will have the right to present argument on this point. Even though Congress has the power to regulate the jurisdiction of the inferior federal courts, Keene Corp. v. United States, 508 U.S. 200, 207 (1993), citing Finley v. United States, 490 U.S. 545, 548 (1989), under Klein it cannot take away the power of the court in a particular case to preserve the status quo while it ponders these weighty questions.

This conclusion makes it unnecessary for us to reach the alternate constitutional grounds on which the prisoners have relied. We emphasize again that our ruling is a narrow one. It does not prevent the state from asking that the prospective relief be stayed while an immediate termination motion is pending. Furthermore, in Berwanger this court has upheld the constitutionality of the heart of this part of the PLRA, which is the (b)(2) right to immediate termination. Here, we hold only that 18 U.S.C. §3626(e)(2) is unconstitutional insofar as it is a legislatively commanded, self-executing stay of an existing court order that requires remedies in a prison conditions case. The district judge did not abuse his discretion at the time he issued his order refusing to stay prospective relief in this case, while he considered the state's petition for immediate termination of the decree governing the Pendleton Correctional Facility.³ On the understanding that the PLRA requires the district court to take every possible measure to expedite its final ruling on the state's motion, and if need be to reconsider the propriety of an interim stay, the order of the district court is AFFIRMED.

FOOTNOTES

¹Most of the statutes cited by the dissent operate in precisely this fashion. See post at 25-28. Thus, for example, a party who wishes to be exempted from the automatic stay in bankruptcy need only petition the bankruptcy court for relief, and that court can lift the stay. Similarly, the Speedy Trial Act contains numerous safeguards that allow a district court to modify the time limits when good cause exists to do so. If these statutes offer good analogies to §3626(e)(2), then logically the dissent should agree with the position of the Sixth Circuit: the automatic stay of (e)(2) goes into effect, but the district court is free to annul the stay on traditional equitable grounds. As the Sixth Circuit held, under that reading there plainly would be no constitutional flaw in the structure of the system.

²At least one of the other examples cited by the dissent has been construed the same way, namely, the statute governing appeals by persons incarcerated for contempt of a grand jury, 28 U.S.C. §1826(b), see In the Matter of a Witness Before the Special October 1981 Grand Jury, 722 F.2d 349, 353 (7th Cir. 1983).

³Although the Sixth Circuit and this panel have arrived at the same ultimate conclusion- -namely, that the district courts retain the power notwithstanding §3626(e)(2) either to stay prospective relief pending adjudication of a §3626(b)(2) motion or to refuse a stay- -the reasoning underlying our respective decisions is sufficiently at odds that we have circulated this opinion to the full court under Circuit Rule 40(e). A majority of the judges in regular active service did not wish to hear this case en banc. Chief Judge Posner and Circuit Judges Easterbrook and Manion voted to rehear en banc.

EASTERBROOK, Circuit Judge, with whom POSNER, Chief Judge, and MANION, Circuit Judge, join, dissenting from the denial of rehearing en banc. The panel holds an Act of Congress unconstitutional- -and not some musty statute overtaken by a change of constitutional doctrine, but the flywheel of recent legislation. Moreover, the panel's approach is not supported by either the text of the Constitution or any doctrine developed by the Supreme Court, and if the panel is right then many other important statutes and rules are unconstitutional. I agree with the panel that §3626(e)(2) cannot bear the reading given it by Hadix v. Johnson, 144 F.3d 925 (6th Cir. 1998). Congress set a deadline; Hadix turns it into mush. The creation of a conflict is justified. But we should not then use constitutional grounds to knock §3626(e)(2) out of commission. Judge Norris's separate opinion in Hadix, 144 F.3d at 950-52, rightly concludes that this statute is within Congress' power.

The Prison Litigation Reform Act substantially changes the criteria that permit a federal court to take over the management of a prison. Congress directed courts to apply the new criteria to existing decrees as well as to future ones. That change properly may be applied to ongoing relief. So we hold today in Berwanger v. Cottey, No. 98-3107, following the lead of many other courts. See Inmates of Suffolk County Jail v. Rouse, 129 F.3d 649 (1st Cir. 1997); Benjamin v. Jacobson, 1999 U.S. App. LEXIS 5130 (2d Cir. Mar. 23, 1999) (en banc); Imprisoned Citizens Union v. Ridge, 169 F.3d 178 (3d Cir. 1999); Plyler v. Moore, 100 F.3d 365 (4th Cir. 1996); Hadix v. Johnson, 133 F.3d 940 (6th Cir. 1998); Gavin v. Branstad, 122 F.3d 1081 (8th Cir. 1997); Dougan v. Singletary, 129 F.3d 1424 (11th Cir. 1997). The only contrary appellate decision, Taylor v. United States, 143 F.3d 1178 (9th Cir. 1998), has been withdrawn on the grant of rehearing en banc, 158 F.3d 1059 (9th Cir. 1998).

Many judges are reluctant to undo their own handiwork and may share the view of the district judge in Pasadena City Board of Education v. Spangler, 427 U.S. 424, 433 (1976), that relief should continue for the judge's lifetime. Congress therefore designed a mechanism to ensure the application of the new rules to the stock of existing decrees- -and to facilitate reexamination even of post-PLRA orders, so that state and local governments may regain control of their institutions once an injunction has achieved its purpose of correcting violations of federal law. See, e.g., Missouri v. Jenkins, 515 U.S. 70, 102 (1995); People Who Care v. Rockford Board of Education, No. 98-1056 (7th Cir. Mar. 19, 1999), slip op. 11-12.

Once a year has passed since the last time the judge addressed the subject (or two years since the decree's entry or the PLRA's enactment) any party may file a motion to terminate relief. 18 U.S.C. §3626(b)(1). The decree then must be terminated, §3626(b)(2), unless the judge determines "that prospective relief remains necessary to correct a current and ongoing violation of the Federal right, extends no further than necessary to correct the violation of the Federal right, and that the prospective relief is narrowly drawn and the least intrusive means to correct the violation." 18 U.S.C. §3626(b)(3). Deadlines are essential lest these rules prove nugatory, so §3626(e)(l) requires the court to decide "promptly", and §3626(e)(2) quantifies "prompt" as within 30 days, with a possible extension to 90 under §3626(e)(3). Subsection (e)(2) calls for a stay rather than termination of the decree. Like the automatic stay in bankruptcy law, 11 U.S.C. §362, subsection (e)(2) does not cancel any judicial decision; it simply stays its effectiveness until the judge renders a decision on the merits. Subsection (e)(2) thus gives the judge a reason to rule promptly. Our case shows the need for such an incentive. Having declared subsection (e)(2) to be unconstitutional, the district judge proceeded to ignore subsection (e)(1). Defendants sought termination of the decree by a motion in June 1997. Almost two years have passed, but the district judge has yet to take a single step toward acting on this request- -and the last word of the panel's opinion is "affirmed." A process that is supposed to be rapid drags on with no end in sight.

According to the panel, subsection (e)(2) is unconstitutional because it "operates directly on the internal adjudication of a case in federal court." Slip op. 12. This is not an accurate description of the statute. Although subsection (e)(1) tells courts how to behave ("promptly rule"), subsection (e)(2) does not. Like the automatic stay in bankruptcy, this statute tells the parties whether they can take advantage of a judgment; it does not tell judges when, how, or what to do, but specifies what happens if the judge does not act. If 30 days pass without action, prospective relief is automatically stayed. The judge can devote his time to the criminal docket or whatever he deems more pressing than prison-reform litigation. But even if we understand §3626(e)(2) as affecting the court's allocation of time (which it will do indirectly; that's its point), why is this a constitutional problem?

Article III establishes three safeguards of judicial independence: tenure of office, protection against financial penalties, and the rule (an implication of establishing a "judicial Power") that final judgments must be carried out. See Plaut v. Spendthrift Farm, Inc., 514 U.S. 211 (1995). An automatic stay following 30 days of judicial inaction does not undermine any of these. The judge is 100% in charge. Stays do not conflict with tenure, salary protection, or respect for judgments. Bankruptcy practice shows this; often the principal reason for filing is to suspend the effectiveness of a judgment (such as a judgment of foreclosure and sale) pending further decision. Consider, too, Fed.R.Civ.P. 65(b), which limits a temporary restraining order to 10 days (with a single extension to 20). Just as subsection (e)(2) causes an injunction to lapse unless the judge makes findings within 30 (or 90) days, so Rule 65(b) causes an injunction to lapse unless the judge makes findings within 10 (or 20) days. No one thinks that Rule 65(b) is an unconstitutional intrusion on the way judges manage their business; instead it protects defendants against unwarranted judicial interference. Just so with §3626(e)(2): It ensures that state and local governments are not burdened by federal control of their institutions for longer than is necessary.

My colleagues on the panel treat Rule 65(b) as justified by the need to limit ex parte orders, and this is indeed good support for the rule- -but support that is unrelated to Article III. It mixes up distinct issues to distinguish Rule 65(b) from §3626(e)(2) by pointing to the rights of litigants. If Congress should provide that injunctions in prison litigation expire five minutes after the court receives a motion, this would indeed be unconstitutional, but not because of anything in Article III. The problem with my hypothetical statute would be the due process clause, which entitles litigants to a meaningful opportunity to be heard before a final decision. Section 3626(e)(2) is not problematic under the due process clause. A stay is not a final decision (it is more like a TRO, which may issue ex parte), and 30 days is adequate for the litigants to be heard (just as the 20-day window under Rule 65 affords time for a hearing). The panel's concern about defendants who drag their heels in an effort to prevent the judge from reaching a decision (slip op. 12-13) has nothing to do with Article III and the court's internal operations, and everything to do with the due process rights of the litigants. The possibility of foot-dragging under §3626(e)(2) is less serious than under Rule 65(b), for a defendant's barricade of the plaintiff's access to information would be "good cause" to postpone the automatic stay until 90 days under subsection (e)(3), and this extra time makes the tactic less likely to succeed than when the defendant need stall for only 20 days. Courts can foil delaying maneuvers by imposing sanctions on parties that fail to cooperate in discovery, see Fed.R.Civ.P. 37, and by drawing adverse inferences about missing evidence. The inference then could support a finding under §3626(b)(3). None of this, however, has anything to do with Article III or the separation of powers.

Someone who thinks that the PLRA is an intrusion on the powers of the judiciary should point not to §3626(e)(2), which just goads judges to get a move on, but to §3626(b)(2), which calls for the "immediate termination" of injunctive relief on motion of a party. Yet we hold in Berwanger- -and the panel in French agrees, slip op. 7- -that subsection (b)(2) is constitutional, for Congress may require courts to revisit prospective relief when it changes generally applicable rules of law. See Robertson v. Seattle Audubon Society, 503 U.S. 429 (1992); Rufo v. Inmates of Suffolk County Jail, 502 U.S. 367 (1992); System Federation No. 91 v. Wright, 364 U.S. 642 (1961). The PLRA creates new substantive requirements, and the legislature may require all ongoing relief- -whether based on pre-PLRA or post-PLRA decrees- -to conform. If subsection (b)(2) is constitutional, then Congress must be entitled to require its effective implementation, which is all subsection (e)(2) does.

Federal judges have no constitutional power to frustrate through delay the implementation of valid legislation. I am not aware of any decision by the Supreme Court holding, or even suggesting, that statutes requiring judges to adjudicate with dispatch pose constitutional problems. Many laws, of unquestioned validity, require the President and his subordinates to meet deadlines for action; the Judicial Branch has no more freedom from time pressure than does the Executive Branch. If the separation of powers protects judges from time limits, it protects the President too. The panel is unwilling to extend any such protection to the President; hundreds if not thousands of laws would be swept away. But as a matter of constitutional language and structure, the Judicial and Executive Branches are identically situated with respect to deadlines that affect the sequence in which they handle business.

If there is a constitutionally based right of independence in the administration of judicial business, it invalidates 28 U.S.C. §§144 and 455, which prescribe judicial conduct much more directly than does §3626(e) (2). Until 1911, when §144 was enacted, rules of disqualification were based on the common law. A formal Code of Conduct for United States Judges was first adopted in 1973. Not until 1974 was there any requirement that federal judges refrain from sitting when their impartiality might reasonably be questioned. 28 U.S.C. §455(a). Thoroughgoing application of the panel's approach would annul these statutes and return all ethical standards to judicial hands. But if a legislature has the constitutional authority to strip judges of the power to render decisions at all (and this is what §144 and §455 do), it has the authority to tell judges to devote priority attention to cases that the legislature deems vital. Statutes specifying procedures for adjudication have been with us since the beginning of the republic (see Act of Sept. 24, 1789, ch. 20, 1 Stat. 73, 83; Act of Sept. 29, 1789, ch. 21, 1 Stat. 93), something one cannot say about statutory ethics rules. But if the process of adjudication really is independent of legislative control, all procedural rules predating the Rules Enabling Act of 1936- -and all statutes overriding rules promulgated by judges under that law- -must be unconstitutional too.

Interference with judges' allocation of time is only one of the panel's objections to §3626(e)(2). The other rests on United States v. Klein, 80 U.S. (13 Wall.) 128 (1871), which the panel reads as establishing the principle that "Congress does not have the power to impose a rule of decision for pending judicial cases, apart from its power to change the underlying applicable law." Slip op. 17. Whether Klein stands for this proposition is an interesting question, see Seattle Audubon Society, 503 U.S. at 441, but not one we need consider. Subsection (e)(2) does not establish any "rule of decision for pending judicial cases"; the rule of decision comes from §3626(b)(3), a statute that assuredly makes a "change [in] the underlying applicable law." Many judgments governing prison conditions are based on the parties' consent, which may be unrelated to the requirements of federal law. See Firefighters Local 93 v. Cleveland, 478 U.S. 501, 522 (1986): "it is the agreement of the parties, rather than the force of the law upon which the complaint was originally based, that creates the obligations embodied in a consent decree." Some decrees depend on a mixture of state and federal obligations. But after §3626(b)(3) prospective relief is limited strictly to enforcing the requirements of federal law, and the relief must be the minimum necessary to vindicate the federal right. That is a substantial change. All §3626(e)(2) does is stay a decree's effectiveness until the court completes the task of applying the new rule of law. An automatic stay pending final decision no more violates the principle of Klein than does the automatic stay in bankruptcy law, or the automatic termination of a TRO under Rule 65(b). Section 3626(e)(2) requires courts to depart from the approach they take when revisiting other decrees, but "Congress may intervene and guide or control the exercise of the courts' discretion." Weinberger v. Romero-Barcelo, 456 U.S. 305, 313 (1982) (emphasis added).

If §3626(e)(2) is unconstitutional, then a long list of statutes is in jeopardy. I have mentioned a few- -the automatic stay in bankruptcy, rules for judicial disqualification, Fed.R.Civ.P. 65(b)- -but there are more. Here are some members of the endangered species.

1. The Bankruptcy Code provides for an automatic stay of all legal claims; including the enforcement of judgments, 11 U.S.C. §362(a)(2), against the debtor. The stay lasts until a judge comes to a conclusion about the subject, just as with §3626(e)(2). But while, under §3626(e)(2), the judge has 30 or 90 days to act before the stay takes effect, under §362 the judge has zero days. So if §3626(e)(2) violates Article III by giving the judge too little time- -and Klein, by staying a judgment without an intervening change of substantive law- -then §362 is worse on both counts. (Section 362 also is like §3626(e)(2) because it does not require the judge to act; it just deprives a litigant of a judgment's benefits until the judge does act.) One can't logically distinguish §362 by saying that Congress has special powers over bankruptcy or that the automatic stay helps to coordinate the handling of claims. The bankruptcy power is no different from §5 of the 14th Amendment, which undergirds §3626. Legislative powers are pertinent to the question whether Congress may alter private rights; they are not pertinent to the question whether Article III contains a judicial immunity from time limits, or prevents judgments from being subject to stay. Buckley v. Valeo, 424 U.S. 1,132-37 (1976), and its successors reject any argument that Congress' "special competence" over such-and-such a subject enables it to enact a law that violates the separation of powers. That a statute is a sensible way to deal with a problem does not save it against a separation-of-powers challenge. See INS v. Chadha, 462 U.S. 919, 944 (1983); Bowsher v. Synar, 478 U.S. 714, 722, 736 (1986).

2. The Speedy Trial Act requires the court to try the defendant within 70 days or dismiss the indictment. 18 U.S.C. §§3161, 3162(a)(2). This statute, like §3626(e)(2), does not require the court to act, but establishes the consequence of delay. Although courts can grant extensions of the 70-day period, this can't matter to the constitutional analysis. My colleagues' objection to §3626(e)(2) is not that the time is too short- -a claim that could not be reconciled with Rule 65(b)- -but that it establishes a deadline. Any other deadline would be equally offensive to the separation of powers. When the fourth circuit sustained the Speedy Trial Act against an attack based on Article III, it rejected the interpretation of Klein that our panel embraces. United States v. Brainer, 691 F.2d 691, 695-96 (4th Cir. 1982).

3. Appeals by persons incarcerated for contempt of a grand jury must be decided within 30 days. 28 U.S.C. §1826(b). This statute operates directly on the judge rather than on the litigants.

4. The court of appeals must grant or deny, within 30 days, an application for leave to commence a second or successive collateral attack. 28 U.S.C. §§2244(b)(3)(D).

5. Chapter 154 of the Judicial Code (added by the AEDPA) sets multiple time limits in capital cases. 28 U.S.C. §2266(b). These are binding rather than hortatory. Section 2266(b)(4)(B) says that a state may "enforce" these limits by mandamus, and that the court of appeals "shall act on the petition for a writ of mandamus not later than 30 days after the filing of the petition."

6. Pretrial detention following arrest can't exceed 10 days unless the court makes specified findings. 18 U.S.C. §3142(d). This is almost completely parallel to §3626(e)(2)- -and it won't do to distinguish the two by saying that §3142(d) protects the right of suspects to liberty. That would be irrelevant to the independence of the judiciary under Article III, the fulcrum of the panel's opinion, and is at all events no distinction: §3626(e)(2) protects the rights of states to be free from unwarranted injunctions of indefinite duration.

7. The criminal rules contain many time limits: Rule 35(c) caps at 7 days the time to correct an error in a sentence (note that this is an outer limit for judicial action, not for a party to make a motion); Rule 29(c) gives a judge only 7 days to entertain (or extend the time to make) a motion for acquittal, and this time can't be further extended. Carlisle v. United States, 517 U.S. 416 (1996). See also Rules 33 and 34- -more time limits that apply to judges, not just counsel. Rule 45(b) forbids the enlargement of time specified by "Rules 29, 33, 34 and 35, except to the extent and under the conditions stated in them." The district judge in Carlisle treated the 7-day time limit for action in Rule 29(c) as advisory; the Supreme Court disagreed; but if our panel is right, then Carlisle is not only wrong but also unconstitutional, for if the judge fails to act within 7 days defendants may lose valuable rights (and in Carlisle did lose a valuable right- -an acquittal!).

8. Former Fed.R.Crim.P. 35 set a limit of 120 days for action on motions to reduce sentence, and strict enforcement of this limit was endorsed by United States v. Addonizio, 442 U.S. 178 (1979). In United States v. Kimberlin, 776 F.2d 1344 (7th Cir. 1985), and Gaertner v. United States, 763 F.2d 787 (7th Cir. 1985), we held that this time limit was "jurisdictional" and thus prevailed even when the prisoner's motion was timely and only the judge's delay deprived the defendant of an opportunity for a lower sentence. If the panel is right, then Addonizio, Kimberlin, and Gaertner are not only wrong but also unconstitutional.

9. After a defendant is found not guilty by reason of insanity, the judge must hold within 40 days a hearing to determine whether the person is still insane and therefore subject to continuing commitment. 18 U.S.C. §4243(c).

10. Fed.R.Civ.P. 65(b) sets 10-and-20 day limits for temporary restraining orders. The Norris-LaGuardia Act, 29 U.S.C. §107, has a 5-day limit. These time limits force the judiciary to hold hearings and make prompt decisions, else orders expire, in the same way as does §3626(e)(2).

My colleagues' response is that these time limits either have escape hatches despite their absolute language- -a response at war with the panel's conclusion that §3626(e)(2) does not permit departure on equitable grounds, and with the Supreme Court's approach to time limits in Carlisle and Addonizio- -or are unconstitutional themselves. Either way, the scope of the panel's decision is breathtaking. As a practical matter, all of these statutes are gutted. They go by the boards not because of either the Constitution's text or any controlling decision of the Supreme Court, but because the panel has invented a right of the judicial branch to freedom from deadlines. If this does not meet the standard for en banc review, I don't know, what does.