RICHARDSON, Circuit Judge

ARGUED: James W. Craig, THE RODERICK & SOLANGE MACARTHUR JUSTICE CENTER, New Orleans, Louisiana, for Appellants. Margaret Hoehl O’Shea, OFFICE OF THE ATTORNEY GENERAL OF VIRGINIA, Richmond, Virginia, for Appellee. ON BRIEF: Emily M. Washington, THE RODERICK & SOLANGE MACARTHUR JUSTICE CENTER, New Orleans, Louisiana; Christina Bonanni, William C. Miller, PILLSBURY WINTHROP SHAW PITTMAN LLP, Washington, D.C.; David M. Shapiro, Roderick & Solange MacArthur Justice Center, NORTHWESTERN PRITZKER SCHOOL OF LAW, Chicago, Illinois, for Appellants. Mark R. Herring, Attorney General, Victoria N. Pearson, Deputy Attorney General, Toby J. Heytens, Solicitor General, Matthew R. McGuire, Principal Deputy Solicitor General, Michelle S. Kallen, Deputy Solicitor General, OFFICE OF THE ATTORNEY GENERAL OF VIRGINIA, Richmond, Virginia, for Appellee. Douglas J. Peterson, Attorney General, James D. Smith, Solicitor General, David A. Lopez, Deputy Solicitor General, Ryan S. Post, Assistant Attorney General, OFFICE OF THE ATTORNEY GENERAL OF NEBRASKA, Lincoln, Nebraska; Steve Marshall, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF ALABAMA, Montgomery, Alabama; Leslie Rutledge, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF ARKANSAS, Little Rock, Arkansas; Mark Brnovich, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF ARIZONA, Phoenix, Arizona; Pamela Jo Bondi, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF FLORIDA, Tallahassee, Florida; Christopher M. Carr, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF GEORGIA, Atlanta, Georgia; Lawrence G. Wasden, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF IDAHO, Boise, Idaho; Curtis T. Hill, Jr., Attorney General, OFFICE OF THE ATTORNEY GENERAL OF INDIANA, Indianapolis, Indiana; Derek Schmidt, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF KANSAS, Topeka, Kansas; Jeff Landry, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF LOUISIANA, Baton Rouge, Louisiana; Joshua D. Hawley, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF MISSOURI, Jefferson City, Missouri; Adam Paul Laxalt, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF NEVADA, Carson City, Nevada; Mike Hunter, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF OKLAHOMA, Oklahoma City, Oklahoma; Alan Wilson, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF SOUTH CAROLINA, Columbia, South Carolina; Marty J. Jackley, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF SOUTH DAKOTA, Pierre, South Dakota; Ken Paxton, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF TEXAS, Austin, Texas; Sean D. Reyes, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF UTAH, Salt Lake City, Utah; Peter K. Michael, Attorney General, OFFICE OF THE ATTORNEY GENERAL OF WYOMING, Cheyenne, Wyoming, for Amici Curiae.

Before AGEE, KEENAN, and s.

Affirmed by published opinion. Judge Richardson wrote the opinion, in which Judge Agee and Judge Keenan concurred.

RICHARDSON, Circuit Judge:Richard Jordan and Ricky Chase, both Mississippi death-row inmates, have filed a federal lawsuit challenging that state’s lethal-injection procedures under the Eighth Amendment. As part of discovery in that lawsuit, they served a subpoena on the Virginia Department of Corrections ("VDOC") that sought documents and testimony about Virginia’s execution practices. VDOC provided some documents to Jordan and Chase and then moved to quash the subpoena in district court. The district court granted VDOC’s motion, finding that requiring VDOC to provide any further response to the subpoena would impose an undue burden. VDOC urges us to affirm on the merits or, if it would not prevail on the merits, because the subpoena infringes Virginia’s state sovereign immunity.

As we explain below, we need not reach state sovereign immunity given VDOC’s conditional assertion of that defense. Instead, we affirm on the merits. The district court reasonably found that Jordan and Chase did not have a need for further discovery from VDOC, a nonparty, that outweighed the burdens the discovery would impose. The district court thus did not abuse its discretion in quashing the subpoena.

I.

This appeal is ancillary to a civil action pending in the Southern District of Mississippi. Jordan and Chase, the plaintiffs in that action, challenge Mississippi’s method of execution, which they claim will cause them unnecessary physical pain. Mississippi executes prisoners by lethal injection using a three-drug protocol that begins with the administration of an anesthetic. At the start of the plaintiffs’ litigation, that anesthetic was the drug pentobarbital ; after the identity of Mississippi’s pentobarbital supplier was disclosed, the state had to replace pentobarbital with another drug, midazolam. Mississippi’s loss of its pentobarbital supplier was just one instance of a well-known phenomenon in which drug suppliers, once exposed to pressure from activists opposed to the death penalty, refuse to supply drugs to state corrections departments. See generally Glossip v. Gross , ––– U.S. ––––, 135 S.Ct. 2726, 2733–34, 192 L.Ed.2d 761 (2015) (discussing this trend).

Jordan and Chase have mounted several attacks on Mississippi’s execution practices. They argue that Mississippi should use a one-drug protocol that consists of a large dose of a single anesthetic, not a three-drug protocol that uses an anesthetic followed by other drugs. They also claim that midazolam should not be used as the anesthetic in the three-drug protocol, because it is inadequate to prevent pain during execution. Jordan and Chase further challenge Mississippi’s use of drugs prepared by compounding pharmacies that, in their view, lack meaningful regulatory oversight. To prevail in their attacks, Jordan and Chase must show "a feasible and readily implemented alternative method of execution that would significantly reduce a substantial risk of severe pain and that the State has refused to adopt without a legitimate penological reason." Bucklew v. Precythe , ––– U.S. ––––, 139 S.Ct. 1112, 1125, ––– L.Ed.2d –––– (2019) (citing Glossip , 135 S.Ct. at 2737–38 ; Baze v. Rees , 553 U.S. 35, 52, 128 S.Ct. 1520, 170 L.Ed.2d 420 (2008) (plurality opinion)).

Jordan and Chase have tried to meet this burden, in part, by comparing Mississippi’s practices to those of other states using lethal injection. Virginia is one such state. Like Mississippi, Virginia carries out executions using a three-drug protocol that includes compounded midazolam. Like Mississippi, it previously used pentobarbital in place of midazolam. And like Mississippi, Virginia is eager to keep the identity of its drug supplier secret, lest pressure from death-penalty opponents cause the supplier to stop providing the state with drugs for use in executions.

During discovery, Jordan and Chase served a third-party subpoena on VDOC. The subpoena sought both documents and testimony. The document requests covered various topics relating to Virginia’s executions, including how Virginia has obtained its execution drugs. The testimony was sought under Federal Rule of Civil Procedure 30(b)(6), which authorizes parties to take testimony from entities like corporations and government agencies. A Rule 30(b)(6) subpoena must identify the subject matter of the testimony sought; the topics listed in Jordan and Chase’s Rule 30(b)(6) request overlapped substantially with their document requests.

In response to the subpoena, VDOC provided Jordan and Chase with a number of documents, many of which originated in prior litigation about Virginia’s execution practices. The documents included labels and certificates of analysis for Virginia’s execution drugs, a redacted copy of Virginia’s agreement with the compounding pharmacy that serves as Virginia’s drug supplier, a redacted copy of VDOC’s execution manual, and transcripts from evidentiary hearings in the prior litigation. The transcripts included a discussion of an episode in 2015 in which a VDOC official drove to Texas to obtain pentobarbital from Texas officials for use in a Virginia execution. VDOC informed Jordan and Chase’s counsel that it was willing to provide an additional affidavit to "fill in any of the gaps," but that it otherwise objected to the subpoena and would move to quash if the parties could not agree on how to narrow the scope of the subpoena. J.A. 161.

No agreement was reached, and VDOC moved to quash. VDOC argued that the subpoena sought privileged material; unduly burdened Virginia, its drug supplier, and the members of its execution team; and violated Virginia’s state sovereign immunity. VDOC described the subpoena as an overbroad "fishing expedition" into its execution practices, which had limited (if any) relevance to Jordan and Chase’s Mississippi lawsuit. J.A. 69–70. What is more, VDOC argued, the outstanding requests in the subpoena would reveal the identity of Virginia’s supplier of execution drugs. As VDOC pointed out, that information is normally kept confidential: Virginia state law recognizes a qualified privilege for the identity of execution-drug suppliers. Va. Code § 53.1-234. Disclosing this information would not only infringe on the drug supplier’s own interest in confidentiality, VDOC argued, but also restrict Virginia’s ability to carry out lawful executions by chilling suppliers from providing execution drugs.

In response, Jordan and Chase argued that the information they sought was relevant and that disclosure would not impose a burden on VDOC. In their view, they needed discovery from VDOC to show the required "feasible and readily implemented" alternative. Information about Virginia’s execution drugs would be relevant (1) to evaluate Mississippi’s claim that using pentobarbital is not a "feasible" alternative because Mississippi cannot acquire pentobarbital, and (2) to show the feasibility of the one-drug protocol. Jordan and Chase also suggested that Virginia might have evidence that the use of midazolam in the three-drug protocol poses an excessively high risk of suffering. In support of this claim, they pointed to a 2017 execution that Virginia officials carried out using midazolam, during which an unusual 30-minute delay occurred. J.A. 225–26. Jordan and Chase opposed Virginia’s argument that there was a cognizable confidentiality interest in the identity of its drug supplier. And even if there was, they argued, a confidentiality order could protect it.

In a 42-page opinion, the district court granted VDOC’s motion to quash because the requests imposed an undue burden, without reaching the parties’ arguments on privilege and state sovereign immunity. The court found that VDOC’s prompt responses had provided much of the requested information. Additional information was unlikely to be relevant to the key issue: whether there is a safer, feasible alternative to Mississippi’s execution method. Virginia’s method was not really an "alternative" at all. Like Mississippi, Virginia uses a three-drug protocol that includes compounded midazolam. Nor could Virginia’s experience be used to disprove Mississippi’s claim that it could not obtain pentobarbital: VDOC produced documents showing that it, too, could not obtain pentobarbital. The district court also rejected Jordan and Chase’s suggestion that Virginia’s execution practices were flawed. Jordan and Chase’s only evidence was the 30-minute delay during the 2017 execution, which was not compelling proof given Virginia’s well-documented history of successfully using the three-drug protocol. The court also noted that Jordan and Chase had failed to explain what relevant information they could receive from Mississippi officials, casting more doubt on their need for additional discovery from VDOC.

The district court also found that further disclosure would burden Virginia’s ability to obtain lethal-injection drugs. Many of the documents Jordan and Chase sought would reveal the identity of Virginia’s midazolam supplier, either directly or by giving clues that could lead someone to uncover it. Documents provided by VDOC show the sensitivity of this information. In hearing transcripts from the prior litigation, VDOC officials described how hard it was to find a supplier of execution drugs. Only after Virginia passed its confidentiality statute was VDOC able to find a compounding pharmacy willing to supply it with midazolam, and even then only with an agreement promising to protect the supplier’s confidentiality. In addition, the district court noted the pressure that death-penalty opponents place on drug suppliers. So the district court found that disclosing the identity of the drug supplier would inhibit Virginia’s ability to carry out lawful executions. And the district court found that a confidentiality order would not adequately alleviate this burden. The court also found the subpoena overbroad for seeking information dating back to 2010, far longer than reasonably necessary.

Based on all these considerations, the district court found that the subpoena imposed an undue burden and granted VDOC’s motion to quash.

Jordan and Chase timely appeal. Because the court below denied nonparty discovery in an ancillary proceeding where the underlying action is pending in another circuit, we have appellate jurisdiction under the collateral-order doctrine. Nicholas v. Wyndham Int’l, Inc. , 373 F.3d 537, 541 (4th Cir. 2004).

II.

At the outset, we must determine the proper order of decision: do we start with VDOC’s state sovereign immunity argument or the merits? We conclude that we should start with the merits.

Courts must generally decide jurisdictional issues first. That is certainly true of subject matter jurisdiction: courts must always assure themselves of subject matter jurisdiction before reaching the merits, even if the parties have not raised it. See Arbaugh v. Y&H Corp. , 546 U.S. 500, 514, 126 S.Ct. 1235, 163 L.Ed.2d 1097 (2006). It is also true of personal jurisdiction: even though personal jurisdiction may be waived, if it is timely raised, it too takes priority over the merits. Sinochem Int’l Co. v. Malaysia Int’l Shipping Corp. , 549 U.S. 422, 430–31, 127 S.Ct. 1184, 167 L.Ed.2d 15 (2007). The Supreme Court has cautioned against exceptions that allow courts, for reasons of expedience, to sidestep jurisdictional issues. Most notably, in Steel Co. v. Citizens for a Better Environment , 523 U.S. 83, 118 S.Ct. 1003, 140 L.Ed.2d 210 (1998), the Court overturned the practice of exercising "hypothetical jurisdiction" to reach the merits first when the merits question was easy and the jurisdictional question was hard. Id. at 93–94, 118 S.Ct. 1003.

State sovereign immunity should also, at least in principle, have priority over merits issues. It too is a jurisdictional doctrine, although a singular one that is somewhat of a hybrid between subject matter and personal jurisdiction. Wisconsin Dep’t of Corr. v. Schacht , 524 U.S. 381, 394, 118 S.Ct. 2047, 141 L.Ed.2d 364 (1998) (Kennedy, J., concurring); Caleb Nelson, Sovereign Immunity as a Doctrine of Personal Jurisdiction , 115 HARV. L. REV. 1559, 1609–11 (2002). Like other constitutional limitations on our subject matter jurisdiction, "the fundamental principle of sovereign immunity limits the grant of judicial authority," Pennhurst State Sch. & Hosp. v. Halderman , 465 U.S. 89, 98, 104 S.Ct. 900, 79 L.Ed.2d 67 (1984), and is not forfeited by inattention at the trial level, meaning it can be raised for the first time on appeal, Edelman v. Jordan , 415 U.S. 651, 677–78, 94 S.Ct. 1347, 39 L.Ed.2d 662 (1974). Yet state sovereign immunity is unlike subject matter jurisdiction, and like personal jurisdiction, in that it can be affirmatively waived, see Lapides v. Bd. of Regents of Univ. Sys. of Georgia , 535 U.S. 613, 618, 122 S.Ct. 1640, 152 L.Ed.2d 806 (2002), and need not be raised by the court on its own motion, see Patsy v. Board of Regents , 457 U.S. 496, 515–16 n.19, 102 S.Ct. 2557, 73 L.Ed.2d 172 (1982).

The Supreme Court has therefore recognized that sovereign immunity must generally be decided first, but with an exception: courts may instead begin by asking if a statutory cause of action can, under the statute itself, ever be asserted against a state. Vermont Agency of Nat. Res. v. United States ex rel. Stevens , 529 U.S. 765, 779, 120 S.Ct. 1858, 146 L.Ed.2d 836 (2000). There are two reasons for this exception. First, a statute that does not permit suits against a state does not implicate the Eleventh Amendment, and so the statutory question is "logically antecedent" to the constitutional one. Id. Second, the exception is limited to the threshold, categorical question of whether a cause of action exists against a state; it does not permit adjudication of whether the particular facts alleged describe a violation of the statute. For that reason, the exception "does not, as a practical matter, permit the court to pronounce upon any issue, or upon the rights of any person, beyond the issues and persons that would be reached under the Eleventh Amendment inquiry anyway." Id.

Our precedents have identified another exception for cases when a state asserts its sovereign immunity only conditionally. If the state does not "insist" on its sovereign immunity defense, but uses it only as a backstop to prevent a defeat on the merits, we may affirm on the merits alone. Strawser v. Atkins , 290 F.3d 720, 729–30 (4th Cir. 2002) (alternative holding). We have grounded this exception in part in the principle of constitutional avoidance, which counsels against reaching state sovereign immunity where it is unnecessary. Id. at 730. To be sure, one might question whether this exception is in keeping with the spirit of the Supreme Court’s opinions in Steel Co. and Vermont Agency . But we are bound by our precedents, which have concluded that it is. See Constantine v. Rectors & Visitors of George Mason Univ. , 411 F.3d 474, 479–83 (4th Cir. 2005) ; Strawser , 290 F.3d at 729–30 ; see also McClendon v. Georgia Dep’t of Cmty. Health , 261 F.3d 1252, 1258–59 (11th Cir. 2001) (adopting same exception for "conditional" assertions of immunity).

The exception applies to VDOC’s conditional assertion of state sovereign immunity here. At oral argument, VDOC first insisted that we reach state sovereign immunity. Because this Circuit has not addressed whether a subpoena issued against a nonparty state agency—not merely a state official—runs afoul of that state’s sovereign immunity, we ordered supplemental briefing. VDOC then backpedaled in its supplemental brief, withdrawing its prior insistence and stating that it would be content with a victory on the merits. VDOC’s final answer controls, and so the exception applies.

As we explain below, the merits favor VDOC. So we bypass state sovereign immunity.

III.

We now turn to the merits of the motion to quash. District courts enjoy "considerable discretion in overseeing discovery," and we will disturb a district court’s discovery rulings only if we find an abuse of that discretion. Kolon Indus. Inc. v. E.I. DuPont de Nemours & Co. , 748 F.3d 160, 172, 178 (4th Cir. 2014). In applying the abuse-of-discretion standard, we review the district court’s factual conclusions for clear error and its legal conclusions de novo. In re Grand Jury Subpoena , 870 F.3d 312, 316 (4th Cir. 2017). Here, the district court quashed the subpoena after determining it imposed an undue burden. Finding no abuse of discretion, we affirm.

A.

All civil discovery, whether sought from parties or nonparties, is limited in scope by Rule 26(b)(1) in two fundamental ways. First, the matter sought must be "relevant to any party’s claim or defense." Fed. R. Civ. P. 26(b)(1). Relevance is not, on its own, a high bar. There may be a mountain of documents and emails that are relevant in some way to the parties’ dispute, even though much of it is uninteresting or cumulative. Rule 26 therefore imposes another requirement: discovery must also be "proportional to the needs of the case." Id. Proportionality requires courts to consider, among other things, "whether the burden or expense of the proposed discovery outweighs its likely benefit." Id. This relieves parties from the burden of taking unreasonable steps to ferret out every relevant document.

When discovery is sought from nonparties, however, its scope must be limited even more. Nonparties are "strangers" to the litigation, and since they have "no dog in [the] fight," they have "a different set of expectations" from the parties themselves. Cusumano v. Microsoft Corp. , 162 F.3d 708, 717 (1st Cir. 1998). Bystanders should not be drawn into the parties’ dispute without some good reason, even if they have information that falls within the scope of party discovery. For example, a party’s email provider might well possess emails that would be discoverable from the party herself. But unless the email provider can offer important information that cannot be obtained from the party directly, there would be no cause for a subpoena against the provider.

A more demanding variant of the proportionality analysis therefore applies when determining whether, under Rule 45, a subpoena issued against a nonparty "subjects a person to undue burden" and must be quashed or modified. Fed. R. Civ. P. 45(d)(3)(A)(iv). As under Rule 26, the ultimate question is whether the benefits of discovery to the requesting party outweigh the burdens on the recipient. In re Modern Plastics Corp. , 890 F.3d 244, 251 (6th Cir. 2018) ; Citizens Union of N.Y.C. v. Att’y Gen. of N.Y. , 269 F.Supp.3d 124, 138 (S.D.N.Y. 2017). But courts must give the recipient’s nonparty status "special weight," leading to an even more "demanding and sensitive" inquiry than the one governing discovery generally. In re Public Offering PLE Antitrust Litig. , 427 F.3d 49, 53 (1st Cir. 2005).

Without exhaustively listing the considerations that go into this analysis, we note a few that are important here. On the benefit side of the ledger, courts should consider not just the relevance of information sought, but the requesting party’s need for it. See Wiwa v. Royal Dutch Petrol. Co. , 392 F.3d 812, 818 (5th Cir. 2004). The information sought must likely (not just theoretically) have marginal benefit in litigating important issues. (We mean "marginal" in the economic sense that the information must offer some value over and above what the requesting party already has, not in the sense that a mere de minimis benefit will suffice.) Courts should also consider what information is available to the requesting party from other sources. See 9A CHARLES ALAN WRIGHT & ARTHUR R. MILLER, FEDERAL PRACTICE & PROCEDURE § 2463.1, at 501–06 (3d ed. 2008). To that end, the requesting party should be able to explain why it cannot obtain the same information, or comparable information that would also satisfy its needs, from one of the parties to the litigation—or, in appropriate cases, from other third parties that would be more logical targets for the subpoena.

On the burden side, district courts should of course consider the dollars-and-cents costs associated with a large and demanding document production. But there are other cognizable burdens as well. For example, a subpoena may impose a burden by invading privacy or confidentiality interests. See Modern Plastics , 890 F.3d at 251–52 ; In re Missouri Dep’t of Corr. , 839 F.3d 732, 736–37 (8th Cir. 2016) ; Fappiano v. City of New York , 640 F. App'x 115, 121 (2d Cir. 2016). Courts may consider the interests of the recipient of the subpoena, as well as others who might be affected. The text of Rule 45 makes that clear, encompassing burdens on any "person," not just the recipient of the subpoena. Fed. R. Civ. P. 45(d)(3)(A)(iv). For example, if a subpoena seeks information from a business about its customers, it may implicate the business’s interest in protecting competitively sensitive information, as well as the customers’ interest in protecting their privacy, see Modern Plastics , 890 F.3d at 251–52. Another type of burden arises when a subpoena is overbroad—that is, when it seeks information beyond what the requesting party reasonably requires. See Wiwa , 392 F.3d at 818. A nonparty should not have to do the work of tailoring a subpoena to what the requesting party needs; the requesting party should have done that before serving it.

B.

These factors appropriately led the district court to conclude that any further production by VDOC would impose an undue burden. In framing its analysis, the district court properly considered the fact that VDOC had produced responsive documents. That production addressed, at least to some degree, most of the document requests and deposition topics. We see no error in that finding. Thus, the issue was whether providing the documents and testimony that remained outstanding—assuming they satisfied Rule 26 ’s threshold relevance requirement—would create an undue burden.

We begin with the asserted need for the information. Jordan and Chase advanced three theories below for how the documents might bolster their case. First, they claimed that information about actual or potential drug suppliers might undercut Mississippi’s defense that it could no longer acquire pentobarbital. Second, they argued, VDOC might have internal documents discussing the feasibility of a one-drug protocol. Third, because Virginia’s execution protocol supposedly has flaws (as revealed by alleged missteps during a 2017 execution), VDOC’s documents might show that Mississippi’s similar three-drug protocol poses a substantial risk of serious harm. On appeal they add a fourth, alternative theory: if Virginia’s execution practices are in fact sound, they can be held up against Mississippi’s practices to show that the latter are flawed.

The district court rejected all of these theories, finding it unlikely that VDOC would have useful information under any of them. We find its conclusions reasonable. First, VDOC had already provided documents showing that it, like Mississippi, could not obtain pentobarbital and had been unable to do so for several years. Additional documents confirming this fact would be of little value. Second, VDOC was an unlikely source of helpful information about a single-drug protocol, because VDOC did not use that protocol and never had. Third, the district court reasonably rejected Jordan and Chase’s factual claim that Virginia’s execution practices were flawed, citing Virginia’s long history of using a three-drug protocol successfully. The district court also anticipated Jordan and Chase’s alternative argument that VDOC might have helpful evidence about procedural safeguards that Mississippi lacked. The court noted that VDOC had produced its execution manual and testimony describing drug storage and testing; Jordan and Chase failed to explain what other information they needed.

Even if VDOC might have additional, helpful evidence on any of these fronts, Jordan and Chase have never explained why VDOC is an appropriate source given the alternatives. As the district court noted, Jordan and Chase "have been particularly opaque as to why the VDOC, a nonparty to the underlying action, is a better source of information than the Mississippi DOC." J.A. 26. Even on appeal, they have provided little insight into what information Mississippi has provided or why it is inadequate.

There are also obvious third-party alternatives. Evidence showing the feasibility and safety of a single-drug protocol could be sought more readily from states that actually use single-drug protocols. Similarly, information about the episode in which Texas officials provided pentobarbital could be sought from them. Jordan and Chase’s failure to address these issues is a mystery, given that they have served subpoenas seeking similar discovery from other state governments.

Jordan and Chase try to bolster their arguments by analogizing to products-liability cases in which plaintiffs have taken third-party discovery from the defendants’ competitors about products with safer alternative designs. That information was arguably relevant because, in at least some jurisdictions, a plaintiff bringing a design-defect claim must show that a reasonable alternative exists. See Restatement (Third) of Torts: Products Liability § 2 (Am. Law Inst. 1988). But this analogy presupposes that Virginia’s execution practices represent a relevant, safer "alternative." In fact, Virginia and Mississippi use similar three-drug execution protocols, and Jordan and Chase have provided nothing, apart from conjecture, to suggest that Virginia’s practices are different or safer. Even if Virginia’s execution practices were marginally better than Mississippi’s, that alone would not make Virginia a relevant point of comparison. Jordan and Chase must prove there is a feasible alternative procedure that "significantly reduce[s] a substantial risk of severe pain," a standard not satisfied "merely by showing a slightly or marginally safer alternative." Baze v. Rees , 553 U.S. 35, 51–52, 128 S.Ct. 1520, 170 L.Ed.2d 420 (2008) (plurality opinion) (emphases added); see also Bucklew v. Precythe ,––– U.S. ––––, 139 S.Ct. 1112, 1130, ––– L.Ed.2d –––– (2019) (holding that "the difference must be clear and considerable").

The analogy to the products-liability context—like many of Jordan and Chase’s arguments—also operates at an excessively high level of generality. It is not tethered close enough to the facts to show why these plaintiffs in this case should get additional discovery from VDOC . When framed at such a high level of abstraction, the argument proves too much. It would, if accepted, entitle every plaintiff in every death-penalty challenge to extensive discovery from every state that conducts executions. Such freewheeling nonparty discovery would improperly turn district courts into "boards of inquiry charged with determining ‘best practices’ for executions," Baze , 553 U.S. at 51, 128 S.Ct. 1520 (plurality opinion), which they are not.

On the other side of the ledger, the district court properly considered two burdens that these requests would impose. First, the district court found the subpoena overbroad for seeking information dating back to 2010. The court concluded that the documents provided already offered "a full picture of how the VDOC currently carries out an execution," J.A. 44–45, and that the older information would not "advance [Jordan and Chase’s] underlying Eighth Amendment claim," J.A. 45. By failing to tailor their subpoena to their needs, Jordan and Chase imposed a burden on VDOC, which had to do the tailoring itself. We find no error in the district court’s conclusion on this point, which Jordan and Chase hardly challenge on appeal.

Second, the district court reasonably found that both VDOC and its drug supplier would be burdened by the disclosure of the supplier’s identity, which most of the outstanding requests sought to uncover. This would harm the supplier’s own confidentiality interest, one recognized by Virginia law. It would also impede Virginia’s ability to carry out executions by chilling Virginia’s current drug supplier, as well as potential future suppliers, from providing drugs for executions. Here, the district court’s findings properly rested on Virginia’s history of difficulties in finding a drug supplier, as well as the fact that death-penalty opponents have pressured drug suppliers in an effort to halt executions. In recognizing that both states and execution-drug suppliers have a legitimate interest in keeping suppliers’ identities confidential, we follow several of our sister circuits. See In re Missouri Dep’t of Corr. , 839 F.3d at 736–37 (reversing, on writ of mandamus, district court’s decision to allow discovery of execution-drug supplier); Arthur v. Commissioner, Alabama Dep’t of Corr. , 840 F.3d 1268, 1304–05 (11th Cir. 2016) (upholding district court’s decision to require disclosure of general facts about drug suppliers, but not their names); In re Ohio Execution Protocol Litig. , 845 F.3d 231, 237–38 (6th Cir. 2016) (upholding district court grant of protective order denying plaintiff discovery of drug suppliers’ identities); In re Lombardi , 741 F.3d 888, 894–96 (8th Cir. 2014) (reversing, on writ of mandamus, district court decision to allow discovery of identities of drug supplier, testing lab, and prescribing physician).

Jordan and Chase respond, in part, by arguing that the cost of compliance would be low because the documents they seek are not so voluminous. But as we have already explained, the expense of collecting, reviewing, and producing documents is not the only burden cognizable under Rule 45. The overbreadth and confidentiality concerns identified by the district court also represent cognizable burdens under Rule 45.

In sum, the district court did not abuse its discretion in finding that Jordan and Chase had little, if any, demonstrated need for the additional documents requested; that they failed to explain why the same or similar information could not be had from better, alternative sources; and that the subpoena imposed cognizable burdens on VDOC and its drug supplier—both nonparties to the litigation. And those considerations, taken together, supported the district court’s finding that the burdens of further compliance with the subpoena outweighed its benefits.

C.

Jordan and Chase alternatively argue that, assuming the district court properly found an undue burden grounded in the need to protect the drug supplier’s identity, it erred by quashing the subpoena outright. They claim the district court instead should have permitted the requested discovery subject to an order requiring the parties to keep the materials confidential. And disclosure subject to a confidentiality order, if feasible, is generally preferable to an outright denial of discovery. Yet the district court considered the issue and concluded that such an order would not have adequately protected the confidentiality interests at stake. We find no abuse of discretion in that decision.

As courts have recognized, sometimes "even the most rigorous efforts of the recipient of [sensitive] information to preserve confidentiality in compliance with the provisions of such a protective order may not prevent inadvertent compromise." In re Deutsche Bank Tr. Co. Americas , 605 F.3d 1373, 1378 (Fed. Cir. 2010). "[I]t is very difficult for the human mind to compartmentalize and selectively suppress information once learned, no matter how well-intentioned the effort may be to do so." Id. (alteration in original) (quoting FTC v. Exxon Corp. , 636 F.2d 1336, 1350 (D.C. Cir. 1980) ). In particular, " ‘extremely potent’ confidential information" may be "of such a nature that it would be ‘humanly impossible’ to control its inadvertent disclosure." Id. (quoting U.S. Steel Co. v. United States , 730 F.2d 1465, 1467 (Fed. Cir. 1984) ).

The information in question—the identity of Virginia’s supplier of execution drugs—is certainly potent. The provision of execution drugs has become a flash point in the ongoing debate over the death penalty, which has long been a contentious issue. Jordan and Chase’s lawyers, advocates trying to prevent their clients’ executions, might find it challenging to keep that information confidential while adhering to their duty of zealous representation. Broad disclosure could easily result from the very act of litigating their case, perhaps from counsel’s investigation of the supplier, or perhaps from the introduction of evidence about the supplier at trial. See In re Missouri Dep’t of Corr. , 839 F.3d at 737. There is also good reason to think that a confidentiality order would not serve its intended purpose. The goal of the order would be to avoid a chilling effect on Virginia’s drug supplier and other, potential suppliers. Yet a drug supplier would probably take little comfort in knowing that disclosure was limited to death-row inmates and their lawyers.

Based on these considerations, the district court reasonably concluded that a confidentiality order would not be up to the task of protecting the interests of Virginia and its drug supplier. The district court therefore did not abuse its discretion in quashing the subpoena outright.

D.

Jordan and Chase make one final argument. They claim that, even if the district court properly quashed their document requests, it should have permitted them to take deposition testimony under Rule 30(b)(6). The same undue-burden standard "applies to both document and testimonial subpoenas." Watts v. SEC , 482 F.3d 501, 508 (D.C. Cir. 2007). Thus, the question is whether Jordan and Chase’s need for a Rule 30(b)(6) deposition outweighs the burden and expense of having one, given the documents VDOC already provided.

There are two possible arguments in support of the Rule 30(b)(6) deposition. The first is that a deposition would provide additional useful information beyond what the documents themselves provide. Yet Jordan and Chase have simply failed to explain what that information might be. That failure is particularly glaring given the unique characteristics of Rule 30(b)(6) depositions. Rule 30(b)(6) deponents testify on behalf of entities, not themselves, and often do so based on careful advance preparation, not personal knowledge. See, e.g. , Ecclesiastes 9:10-11-12, Inc. v. LMC Holding Co. , 497 F.3d 1135, 1146–47 (10th Cir. 2007) ; Brazos River Auth. v. GE Ionics, Inc. , 469 F.3d 416, 433 (5th Cir. 2006). If VDOC’s document production has already provided the substantive information Jordan and Chase are entitled to, then a Rule 30(b)(6) deponent would likely just repeat that same information in some testimonial form.

The second argument is that a Rule 30(b)(6) deposition is necessary to obtain evidence in a form admissible at trial. Here, Jordan and Chase point to the transcripts that VDOC provided. These transcripts capture the testimony of three state employees in prior litigation about Virginia’s execution practices. Jordan and Chase complain that the transcripts may not be admissible in the Mississippi litigation. We find this speculative argument uncompelling. Jordan and Chase do not represent that they intend to use this testimony at trial, or that the defendants would object to its admissibility.

Moreover, Jordan and Chase did not argue in the proceedings below that they needed live testimony to obtain important information or admissible evidence. So the district court did not abuse its discretion by analyzing the document requests and deposition topics together.

IV.

Nonparties faced with civil discovery requests deserve special solicitude. They should not be drawn into the parties’ dispute unless the need to include them outweighs the burdens of doing so, considering their nonparty status. This undue-burden analysis must be conducted based on the concrete facts and issues in the litigation, not on vague generalities or speculation. This district court reasonably determined that Jordan and Chase’s request for more discovery from VDOC was unsupported by a genuine need that outweighed the burdens involved. Because the district court did not abuse its discretion, the order is

AFFIRMED .