SELYA, Circuit Judge.

The management of complex litigation presents challenges that test the mettle of even the most able trial judge. Given the rigors of this task, we have ceded substantial discretion to the district courts with respect to case management decisions. But that discretion, though wide, is not boundless.

In the case at hand, two case management orders crossed this line: an order that totally denied plaintiffs' counsel access to their own clients and an order that prevented the plaintiffs from seeking plainly relevant discovery. Accordingly, we vacate the judgment below and remand for further proceedings.

I. BACKGROUND

The background facts are set out in exegetic detail in a prior opinion of this court and in the opinion of the court below. See Sam M. ex rel. Elliott v. Carcieri, 608 F.3d 77 (1st Cir.2010); Cassie M. ex rel. Irons v. Chafee (Cassie III), 16 F.Supp.3d 33 (D.R.I.2014). We assume the reader's familiarity with those narratives and rehearse here only the events that bear directly on the issues sub judice.

This putative class action was brought on behalf of ten foster children in the custody of the Rhode Island Department of Children, Youth and Families (DCYF). Because the plaintiffs were minors, the initiators of the suit sought to appear as their next friends under Federal Rule of Civil Procedure 17.

The complaint designated as defendants, in their official capacities, a coterie of state officials (collectively, the State). It sought certification of a class of all minor children who were in (or might enter) DCYF custody based on a report or suspicion of abuse or neglect. Invoking 42 U.S.C. § 1983, the complaint prayed for declaratory and injunctive relief on behalf of the named plaintiffs and the putative class.

The gravamen of the complaint was the allegation that DCYF's failings expose foster children in its custody to an unreasonable risk of harm in violation of their substantive due process rights. Relatedly, the complaint averred that the State had failed to comply in various respects with the Adoption Assistance and Child Welfare Act of 1980, 42 U.S.C. §§ 620-628, 670-679a.

The State moved to dismiss the complaint, and the district court obliged. The court's principal rationale was that the proposed next friends did not satisfy Rule 17's requirements.1 See Sam M. ex rel. Elliott v. Carcieri, 610 F.Supp.2d 171, 181-84 (D.R.I.2009). On appeal, we deemed it fitting that foster children be afforded access to a federal forum to seek redress against their custodian. See Sam M., 608 F.3d at 91-92. After evaluating the qualifications of the proposed next friends, we concluded that they were suitable representatives. See id. at 92-94.

On remand, the case was transferred to a different trier. See D.R.I. R. 105(b). The State again moved to dismiss, this time invoking Federal Rule of Civil Procedure 12(b)(6). Although the district court dismissed as moot the claims of five more plaintiffs, it allowed the case to proceed. See Sam M. ex rel. Elliott v. Chafee, 800 F.Supp.2d 363, 389 (D.R.I.2011). The court then invited the parties to submit updated briefing on the plaintiffs' motion for class certification (which had been filed with the complaint). Later, however, the court advised the parties that it would not address class certification until it had decided dispositive motions on the individual claims.

Concerned about the ravages of mootness, the district court allowed the filing of an amended complaint adding five new plaintiffs. The court then denied without prejudice the motion for class certification.

Pretrial discovery frequently proved contentious, taxing the patience of the district judge and the magistrate judge. Although their efforts bordered on the heroic, we eschew a blow-by-blow account and, instead, fast-forward through seven months of discovery to October 23, 2012. On that date — with an array of unfulfilled discovery requests and motions to compel pending — the State moved for a protective order aimed at limiting the scope of discovery. Its principal thesis was that the plaintiffs were entitled only to information pertaining directly to the named plaintiffs. For example, it objected to many of the plaintiffs' discovery requests on the ground that the information sought went beyond the case files of the individual plaintiffs and, therefore, was not relevant.

The district judge referred the motion to the magistrate judge, who allowed the protective order in part. See Cassie M. ex rel. Irons v. Chafee (Cassie I), No. 07-241, slip op. at 3 (D.R.I. Dec. 17, 2012) (unpublished). His rescript explained that, in light of the district judge's plan to address the claims of the individual plaintiffs on summary judgment before addressing class certification, discovery should be limited to information bearing upon those claims. See id. at 2. No separate order was entered delineating the modified scope of discovery, but the magistrate judge directed the parties to confer in an attempt to narrow their disputes in light of his decision. As a result, the plaintiffs provisionally withdrew portions of their motions to compel.

The plaintiffs appealed the protective order to the district judge, who upheld it. See Fed. R. Civ. P. 72(a). Thereafter, the magistrate judge ruled on a number of lingering discovery issues. His decision confirmed that the protective order "effectively precluded [the plaintiffs] from seeking policy or practice discovery." Cassie M. ex rel. Irons v. Chafee (Cassie II), No. 07-241, 2013 WL 785621, at *2 (D.R.I. Mar. 1, 2013). This depiction was consistent with the way in which the plaintiffs had characterized the protective order in their appeal to the district judge.

Another development transpired while the parties were sparring over the scope of discovery. On February 26, 2013, the plaintiffs renewed their motion for class certification. They argued that the district judge's reluctance to decide the class certification issue was causing substantial prejudice because the claims of ten plaintiffs already had become moot and several more plaintiffs would soon age out of DCYF custody. Following procedural skirmishing not relevant here, the district judge again declined to address class certification and allowed the State to delay a response to the class certification motion until fourteen days after any summary judgment decision.

On July 24, 2013, the district judge performed an about-face. At a status conference, the judge voiced concerns about how long it was taking to bring the case to a head. She suggested that the most efficient course of action would be to dispense with dispositive motions and proceed directly to trial. Acting on her own suggestion, the judge declared that she would set the case for trial on the individual claims without awaiting summary judgment motions.

The plaintiffs objected and moved for various kinds of relief. Pertinently, they implored the district judge to postpone any trial, take up the issue of class certification, and reopen discovery so that they could obtain the policy and custom evidence that had been foreclosed by the protective order. The judge rejected these importunings. Citing the fact that the plaintiffs had been able to depose certain DCYF policymakers, she stated that she "believe[d] that the Plaintiffs had an opportunity and have received evidence and discovery on the question of [DCYF's] policies, procedures, [and] customs." In regard to her unwillingness to address class certification, the judge noted that the plaintiffs were required to prove that their constitutional and statutory rights were being violated and reasoned that such a determination could more expeditiously be made in the context of their individual claims before allowing class-wide discovery.

With an unwanted trial looming, plaintiffs' counsel asked the State to facilitate meetings with their clients. The State demurred, making plain that it would not provide any contact information without a court order. On August 29, 2013, the plaintiffs moved to compel the State to allow plaintiffs' counsel and the next friends to meet with the plaintiffs for purposes of trial preparation.2 The State objected.

At the ensuing hearing, the district judge stated that she viewed the motion for access as a veiled attempt to obtain fact discovery beyond the discovery deadline. When the judge asked whether the named plaintiffs would be called as witnesses, counsel replied that they could not make that determination without meeting with their clients. The judge proceeded to deny the motion, effectively preventing plaintiffs' counsel from speaking with their clients in advance of trial.

By the time that the trial commenced on November 12, 2013, the claims of all but two of the named plaintiffs (Danny B. and Cassie M.) had been rendered moot through aging or adoption. When the plaintiffs rested, the State moved for judgment on partial findings. See Fed. R. Civ. P. 52(c). The district judge took the matter under advisement and, after receiving post-trial briefing, concluded that the plaintiffs had presented insufficient evidence to establish that DCYF's policies and customs had either harmed them or exposed them to an unreasonable risk of harm. See Cassie III, 16 F.Supp.3d at 79. The judge likewise concluded that the plaintiffs had failed to carry their burden of proof with respect to their statutory causes of action. See id. at 80. Judgment entered, and this timely appeal ensued.

II. ANALYSIS

The plaintiffs advance a compendium of claims of error. Two of these claims, which relate to pretrial proceedings, are front and center: that the district court (i) prevented the plaintiffs without sufficient reason from meeting with their lawyers and (ii) improvidently pretermitted discovery relating to DCYF's policies and customs. Because we conclude that the district court abused its discretion both when it denied plaintiffs' counsel access to their clients and when it refused to allow discovery essential to the plaintiffs' claims, we are constrained to vacate the judgment. The tale follows.

A. Access to Counsel.

The plaintiffs submit that the district court abused its discretion when it refused to allow their lawyers to meet with them.3 We have not yet had occasion to consider the standard of review applicable to decisions anent a minor's access to her lawyers. Nor do we need to cross that bridge today: the parties agree that we should review this decision for abuse of discretion, and we accept that agreement. See United States v. Ramirez-Rivera, 241 F.3d 37, 40 & n. 4 (1st Cir.2001); cf. Borden v. Paul Revere Life Ins. Co., 935 F.2d 370, 375 (1st Cir.1991) (holding that federal court sitting in diversity may accept parties' reasonable agreement concerning choice of law).

The district court is responsible for safeguarding the interests of those minors who appear as litigants before it. See Noe v. True, 507 F.2d 9, 12 (6th Cir.1974). The situation is complicated where, as here, the minors are foster children who are suing the State. As we previously noted, the State has a structural conflict of interest when it comes to decisions concerning the legal representation of such plaintiffs. See Sam M., 608 F.3d at 88 n. 12. This conflict is exacerbated because plaintiffs' counsel cannot even gain access to their own clients without the State's cooperation.

The claim of error here tests the way in which the district court handled this conflict. A fundamental principle guides our analysis. Civil litigants have a constitutional right, rooted in the Due Process Clause, to retain the services of counsel. See Gray v. New Eng. Tel. & Tel. Co., 792 F.2d 251, 257 (1st Cir.1986); Potashnick v. Port City Constr. Co., 609 F.2d 1101, 1117-18 (5th Cir.1980). This right safeguards a litigant's interest in communicating freely with counsel both in preparation for and during trial. See Doe v. District of Columbia, 697 F.2d 1115, 1119 (D.C.Cir.1983); Potashnick, 609 F.2d at 1118-19. After all, the right to retain counsel would be drained of meaning if a litigant could not speak openly with her lawyer about her case and how best to prosecute it. See Denius v. Dunlap, 209 F.3d 944, 954 (7th Cir.2000); Doe, 697 F.2d at 1119. While this right is not absolute — courts surely can insist upon reasonable rules of practice that affect the lawyer-client relationship — a court must give great weight to this valued interest even in areas committed to its discretion. See, e.g., Doe, 697 F.2d at 1119-20 (requiring protective order limiting counsel's discussion of discovery materials with client to be narrowly drawn); Potashnick, 609 F.2d at 1119 (reversing judgment where court unreasonably barred attorney from speaking to client during breaks in testimony).

Resolving this appeal does not require us to explore all of the circumstances that might justify limiting communications between a minor plaintiff and her lawyer. Suffice it to say that a court may not restrain a litigant's access to counsel without some substantial justification, and any such restraint should be narrowly tailored to respond to the concern that prompted it. See Doe, 697 F.2d at 1120; In re Ti.B., 762 A.2d 20, 29-30 (D.C.2000); cf. Gulf Oil Co. v. Bernard, 452 U.S. 89, 101-02, 101 S.Ct. 2193, 68 L.Ed.2d 693 (1981) (discussing restrictions on communications with potential class members, and requiring that such orders be "carefully drawn" and "limit[] speech as little as possible"). An order that effectively precludes any pretrial communication between a minor and her attorney must satisfy these criteria.

No substantial justification is apparent here. The record reveals no indication that the court below accorded any weight to the plaintiffs' interest in communicating with counsel. Goaded by the State, the court seemed to assume that the next friends' access to counsel was a sufficient substitute for the plaintiffs' access, reasoning that for purposes "of trial strategy sessions, that's what the next friends are for." But this reasoning does not hold water: even if a next friend's access to counsel sometimes may assuage due process concerns, that would not be true when, as in this case, the next friend was also denied access to the client.

The district court gave two other reasons in support of its preclusive ruling. First, the court posited that the motion for access was untimely because it was filed after the close of discovery. This assumed, of course, that the purpose of the requested meetings was to obtain discoverable information. But that assumption was not grounded in anything that plaintiffs' counsel said or wrote. Rather, plaintiffs' counsel advised the court that the purpose of the proposed meetings was to prepare for trial and to determine whether their clients wished to testify.4

At any rate, the plaintiffs' interest in speaking with their counsel plainly persisted beyond the close of discovery. Even though there was a legitimate concern that counsel might obtain and attempt to introduce new information in derogation of the discovery deadline, the district court had the ability to prevent the use of such new information without resorting to a total denial of access. There is no need to use an elephant gun to slay a mouse, and the court could have allowed the meetings to proceed and simply excluded any newly discovered matter if and when the plaintiffs attempted to use it at trial. See, e.g., Fed. R. Civ. P. 16(f); 37(b)(2)(A)(ii).

The district court's second rationale relates to its allowance of a motion by plaintiffs' counsel to have their chosen psychologist interview the plaintiffs in advance of trial. The court theorized that since it was allowing the psychologist to examine the plaintiffs, plaintiffs' counsel would have the benefit of the psychologist's report and would have no need to meet personally with their clients. We do not agree.

The psychological examinations served an entirely different purpose than the requested attorney-client meetings. Those examinations did not provide counsel an opportunity to explain the litigation, determine how the children would fit into a trial strategy, or ascertain the desirability of having them testify. If more were needed — and we doubt that it is — the psychological examinations were videotaped and by court order delivered to opposing counsel. Manifestly, then, any privileged communication was impossible. That the court allowed the psychological examinations to go forward was irrelevant to whether the State was justified in blocking the lawyers' access to their clients.

In an effort to catch lightning in a bottle, the State tries to justify the denial of access on the ground that attorney-client meetings might not have been in the plaintiffs' best interests. But this is pie in the sky; the State offered no evidence that the plaintiffs might suffer any harm from such meetings.5 Dancing around this point, the State says (without meaningful citation to the record) that the district court denied access, at least in part, out of such a concern. But the district judge, who ably articulated her views at every stage of this complex case, neither expressed any such concern nor made any finding that harm was likely to ensue should the lawyers meet with their clients.6 Consequently, we decline the State's self-serving invitation to impute this reasoning to the court below.

The short of the matter is that the denial of access rests on insupportable findings, which are not strengthened by the State's conjectural hypotheses. There is simply not enough here to justify the total denial of a litigant's right to consult with her lawyer. Cf. Sam M., 608 F.3d at 88 n. 12 (disapproving State's attempt to oppose appointment of next friends based on "a general assertion" that other relatives might be suitable representatives).

Let us be perfectly clear. We do not hold that a district court may never place limits on communications between foster children and their lawyers. In this realm, as in others, the court may make orders that are necessary and appropriate to protect the child's interests. See, e.g., Neilson v. Colgate-Palmolive Co., 199 F.3d 642, 652 (2d Cir.1999); Dacanay v. Mendoza, 573 F.2d 1075, 1079 (9th Cir. 1978). Withal, the limitations imposed by any such order must be no broader than necessary to protect the child. See Doe, 697 F.2d at 1120-21; Ti.B., 762 A.2d at 29-30. Viewed through this prism, the total denial of access imposed here cannot withstand scrutiny. We hold, therefore, that the court below abused its discretion when it completely denied the plaintiffs access to their lawyers prior to trial.

This holding leaves only the question of prejudice — and prejudice is manifest. Apart from a dubious suggestion that plaintiffs' retained psychologist was able to testify about topics that attorney-client meetings might have covered, the State has made no serious attempt to explain how the total denial of access to counsel could be harmless.

The right to counsel is a right of the highest order of importance, and keeping a lawyer from meeting with his client before going to trial frustrates that right. Where, as here, a restraint of that magnitude is imposed without substantial justification, we think that prejudice can fairly be presumed. Cf. United States v. Cronic, 466 U.S. 648, 658-59, 104 S.Ct. 2039, 80 L.Ed.2d 657 (1984) (identifying situations implicating the right to counsel where the circumstances "are so likely to prejudice the accused that the cost of litigating their effect in a particular case is unjustified"). It follows that the judgment below must be vacated. See Doe, 697 F.2d at 1121.

B. Protective Order.

Although the judgment below must be vacated, see supra Part II(A), the prospect of further proceedings counsels in favor of resolving the plaintiffs' contention that the district court erred when it barred discovery of DCYF's policies and customs.7 We start with first principles: district courts must be afforded wide latitude in the management of discovery, and appellate review of such matters is correspondingly deferential. We will disturb a district court's discovery ruling "only upon a clear showing of manifest injustice, that is, where the ... order was plainly wrong and resulted in substantial prejudice." Mack v. Great Atl. & Pac. Tea Co., 871 F.2d 179, 186 (1st Cir.1989). Though this sets a high bar, an order limiting the scope of discovery may constitute reversible error when it is sufficiently prejudicial and based upon an incorrect legal standard or a misapplication of law to fact. See Ji v. Bose Corp., 626 F.3d 116, 122 (1st Cir. 2010); Saldana-Sanchez v. Lopez-Gerena, 256 F.3d 1, 8 (1st Cir.2001).

In the case at hand, the plaintiffs seek to impose liability upon official-capacity state defendants under section 1983. As such, their suit is the functional equivalent of a suit against the sovereign. See Will v. Mich. Dep't of State Police, 491 U.S. 58, 71, 109 S.Ct. 2304, 105 L.Ed.2d 45 (1989). In such a suit, it is black letter law that the plaintiffs must prove that a policy or custom of the State contributed to the alleged violations of federal law in order to prevail. See Hafer v. Melo, 502 U.S. 21, 25, 112 S.Ct. 358, 116 L.Ed.2d 301 (1991); Burrell v. Hampshire Cnty., 307 F.3d 1, 7 (1st Cir.2002). Here, moreover, the plaintiffs seek forward-looking injunctive relief rather than damages; thus, their flagship substantive due process claim turns primarily on whether DCYF's policies and customs subject them to an unconstitutional risk of future harm. See DG ex rel. Stricklin v. Devaughn, 594 F.3d 1188, 1197-98 (10th Cir.2010); see also Helling v. McKinney, 509 U.S. 25, 33-34, 113 S.Ct. 2475, 125 L.Ed.2d 22 (1993) (explaining that injunction may issue to remedy constitutional violation without waiting for manifested harm).

Against this backdrop, we examine the protective order. The lack of a separate document embodying the terms of the protective order hampers this examination. We work with what we have.

The magistrate judge made his ruling in a written rescript. See Cassie I, slip op. at 1-3. This rescript described the protective order as limiting discovery to "nonprivileged information that is relevant to the substantive claims of the individual" plaintiffs. Id. at 2. The plaintiffs appealed the protective order, see Fed. R. Civ. P. 72(a), plausibly characterizing it as a complete bar to policy and custom discovery. This characterization was consistent with the arguments that the State had made to the magistrate judge. When asked to describe its vision of how discovery should unfold, the State explained:

Similarly, the State's supporting memorandum argued "that the requested discovery relating to the policies, patterns or practices... is irrelevant to the determination of [the plaintiffs'] substantive claims. The discovery concerning policies, patterns or practices ... should be reserved until after [the plaintiffs'] individual substantive claims are addressed." In sum, the State drew a sharp distinction between discovery pertaining to harm to the individual plaintiffs (which it argued should be allowed) and policy and custom discovery (which it argued should not be allowed).

The district judge upheld the proscription. She concluded that the protective order was in line with her decision to resolve the individual claims before addressing class certification and, thus, was neither clearly erroneous nor contrary to law.

In a subsequent ruling, the magistrate judge confirmed that the plaintiffs had correctly apprehended the scope of the protective order. He made pellucid that the protective order "effectively precluded [the plaintiffs] from seeking policy or practice discovery." Cassie II, 2013 WL 785621, at *2. The magistrate judge's straightforward characterization of his own order dissolved any possible ambiguity concerning the scope of permissible discovery. Cf. Lefkowitz v. Fair, 816 F.2d 17, 22 (1st Cir. 1987) (explaining that "uncertainty as to the meaning and intendment of a district court order can sometimes best be dispelled by deference to the views of the writing judge"). Thus, the protective order foreclosed the plaintiffs from seeking plainly relevant discovery.

This was an abuse of discretion: a district court may not impose discovery restrictions that preclude a suitor from the legitimate pursuit of evidence supporting her cause of action. See Panola Land Buyers Ass'n v. Shuman, 762 F.2d 1550, 1559-60 (11th Cir.1985) (finding abuse of discretion when court limited discovery to single issue though other issues were relevant); Trevino v. Celanese Corp., 701 F.2d 397, 405-06 (5th Cir.1983) (vacating protective order that barred highly relevant discovery based on court's misapprehension of nature of plaintiff's claims). Nor can it be doubted that the denial of discovery was prejudicial. The protective order forced the plaintiffs to attempt to prove their substantive claims without essential evidence. Given the applicable legal standard, their attempt was destined to fail.

That is exactly what happened: the district judge dismissed the substantive due process claim because she concluded that "[t]here was ... no evidence to establish that DCYF's current policies and practices — or any deliberate disregard of such policies and practices — resulted in harm to the [plaintiffs] or that the [plaintiffs] were subjected to unreasonable risk while in DCYF's care." Cassie III, 16 F.Supp.3d at 79. It is, therefore, nose-on-the-face plain that the adverse decision at trial rested in substantial part on the plaintiffs' failure to adduce precisely the sort of evidence that the protective order prevented them from discovering.

The State offers an interleaved series of arguments in an attempt to shore up the protective order. These arguments are unpersuasive.

The State begins with a suggestion that any failure to obtain policy and custom evidence is the plaintiffs' fault. It maintains that the magistrate judge's decision left the door open for policy and custom discovery, and the plaintiffs — had they elected to do so — could have walked through that portal. The State reaches this hopeful conclusion, however, by selectively parsing the record.

The State's argument emphasizes a statement in the magistrate judge's December 17, 2012 rescript to the effect that "a reasonable range of discovery focused on the [plaintiffs] and the various `system failures' [they have] alleged" would be permitted. Cassie I, slip op. at 3. This language arguably gave rise to a latent ambiguity about the availability of custom and policy discovery. But any ambiguity created by this language was dispelled by how the parties treated the protective order and by the magistrate judge's clarification. See Cassie II, 2013 WL 785621, at *2 (vouchsafing that the protective order "effectively precluded [the plaintiffs] from seeking policy or practice discovery"). The State's optimistic reading of the protective order simply cannot be reconciled with either the parties' actions or the magistrate judge's construction of his own ruling.

In a related vein, the State asserts that it produced ample policy and custom evidence during discovery and that the plaintiffs actually submitted some policy and custom evidence at trial. Building on this foundation, the State further asserts that the protective order worked no prejudice.

There is a grain of truth to these assertions. After all, the plaintiffs did introduce some policy and custom evidence at trial. But there is no way to tell either how much policy and custom evidence somehow came to the plaintiffs' attention or — perhaps more important — how much policy and custom evidence was withheld during discovery. The State has not identified any practical method for allowing us to explore that void, and there is no carefully drawn protective order upon which we can rely. The only thing about which we can be certain is that policy and custom discovery was completely off limits after the magistrate judge issued the protective order.8

Finally, the State exhorts us to affirm the protective order on the ground that the plaintiffs failed to specify what policy and custom information they were seeking when they moved to reopen discovery prior to trial. Its exhortation invokes our admonition, made in the context of Federal Rule of Civil Procedure 56(d), that a party seeking additional discovery must "articulate[] a plausible basis for the... belief that previously undisclosed or undocumented facts exist, that those facts can be secured by further discovery, and that, if obtained, there is some credible prospect that the new evidence will create a trialworthy issue." Mass. Sch. of Law at Andover, Inc. v. Am. Bar Ass'n, 142 F.3d 26, 44 (1st Cir.1998). But there was no Rule 56(d) motion here — indeed, the district judge directed the parties to forgo summary judgment practice — and the comparison that the State seeks to make is apples to oranges.

Unlike a typical Rule 56(d) motion, the plaintiffs' motion involved discovery that previously had been denied by court order. A party whose timely discovery request has been denied cannot reasonably be expected to describe precisely the information responsive to that request that may have been withheld from her. It follows that, in the circumstances of this case, the plaintiffs' motion seeking to reopen discovery was sufficient.

We add a caveat. Our opinion should not be read to suggest that the plaintiffs are entitled to the full range of policy and custom discovery that they sought in the district court. Both the magistrate judge and the district judge expressed legitimate concerns about the breadth of the plaintiffs' discovery requests and the State's plaint that complying with those requests could prove unduly burdensome. But a categorical preclusion of all policy and custom discovery was an over-the-top response to those concerns. On remand, the district court, in its sound discretion, may balance competing considerations of relevance and burdensomeness, and place reasonable limits on the scope of the remaining discovery. See, e.g., Gill v. Gulfstream Park Racing Ass'n, Inc., 399 F.3d 391, 402-03 (1st Cir.2005).

C. Class Certification.

Throughout the proceedings below, the plaintiffs beseeched the district court to certify a plaintiff class sooner rather than later. The district court, for reasons of judicial economy, initially delayed consideration of class certification pending resolution of anticipated motions for summary judgment. But the court then shifted gears: it scrapped any thought of dispositive motions and declared that it would not rule on class certification until after trial. When it later ruled for the State midway through trial, class certification became a dead letter.

The plaintiffs argue that the district court's handling of this issue constituted an abuse of discretion. They submit that the court should have addressed class certification early in the proceedings; that the specter of mootness (through the collateral effects of aging and adoption) made time of the essence; and that, in all events, it was inappropriate to defer a ruling on class certification until after trial on the merits.

Because this case must be remanded for further proceedings, we do not decide the class certification issue. We do, however, offer some general guidance to the district court.

The length of time that a case is pending is not the sole determinant of when a class certification decision should be made. The Civil Rules speak of the need to address class certification at "an early practicable time." Fed. R. Civ. P. 23(c)(1)(A). The word "practicable" imports some leeway in determining the timing of such a decision. See Howe v. Townsend (In re Pharm. Indus. Average Wholesale Price Litig.), 588 F.3d 24, 40 (1st Cir.2009) (explaining that Rule 23(c)(1)(A) gives a court "flexibility to wait to certify the class until the court feels it understands the case and the issues it raises"); Wright v. Schock, 742 F.2d 541, 543 (9th Cir.1984) (similar).

Class certification decisions are context-specific, and each case must be viewed in terms of its own facts. As a general matter, however, Rule 23 permits a district court, in appropriate circumstances, to defer the issue of class certification until after disposing of summary judgment motions. See, e.g., Toben v. Bridgestone Retail Operations, LLC, 751 F.3d 888, 896 (8th Cir.2014); Curtin v. United Airlines, Inc., 275 F.3d 88, 93 (D.C.Cir.2001); Cowen v. Bank United of Tex., 70 F.3d 937, 941 (7th Cir.1995); Wright, 742 F.2d at 543-44. In such a situation, consideration of summary judgment motions is likely to furnish the court the information that it needs to "understand[] the case and the issues it raises." Pharm. Indus., 588 F.3d at 40.

Notwithstanding this flexibility, we are aware of no precedent authorizing a district court, over objection, to conduct a full-blown trial on the merits without pausing to take up a timely motion for class certification. The reason for this lack of precedent seems obvious: at best, trying the individual claims first may prove inefficient; at worst, doing so may create substantial prejudice. See Paxton v. Union Nat'l Bank, 688 F.2d 552, 558-59 (8th Cir.1982); Stastny v. S. Bell Tel. & Tel. Co., 628 F.2d 267, 275 (4th Cir.1980); see also Rodriguez v. Banco Cent., 790 F.2d 172, 175 (1st Cir.1986). The bottom line is that staging a case in this manner puts the cart before the mule.

With these considerations in mind, we counsel the district court that, upon completion of discovery in this case, the parties should be given an opportunity to file dispositive motions. If the case survives summary judgment with the class certification issue still velivolant, the court should decide the motion for class certification.

III. CONCLUSION

We need go no further. The district judge and the magistrate judge, as well as the parties, have labored long and hard over this case, and we are reluctant to push the reset button. But fairness is (and must be) the hallmark of federal-court litigation, and the essence of fairness is the provision of a level playing field. Here, however, the two errors that we have discussed — the total denial of counsel's access to their clients and the imposition of an overly broad protective order — impermissibly tilted the playing field. Consequently, we must vacate the judgment and remand the case for further proceedings consistent with this opinion. Costs are to be taxed in favor of the plaintiffs.

Vacated and Remanded.