No. 24-1975

for the Western District of Michigan at Grand Rapids. No. 1:24-cv-00445 — Robert J. Jonker, District Judge.

Argued: December 10, 2025

Decided and Filed: May 1, 2026

Before: BOGGS, BUSH, and READLER, Circuit Judges.

ARGUED: Richard W. Fanning, Jr., CLARK HILL PLC, Detroit, Michigan, for Appellant. Elise F. Oviedo, NATIONAL LABOR RELATIONS BOARD, Washington, D.C., for Appellee. ON BRIEF: Richard W. Fanning, Jr., Brian D. Shekell, CLARK HILL PLC, Detroit, Michigan, for Appellant. Elise F. Oviedo, Kyle A. Mohr, NATIONAL LABOR RELATIONS BOARD, Washington, D.C., for Appellee.

READLER, J., delivered the opinion of the court in which BUSH, J., concurred. BOGGS, J. (pp. 22 – 34), delivered a separate dissenting opinion.

Page 2READLER, Circuit Judge. Amid a contentious labor election process, Trinity Health declared that it would no longer recognize the union responsible for representing the employees of Trinity’s Grand Haven hospital. While administrative proceedings were pending before the National Labor Relations Board, the Board’s Regional Director petitioned the district court for a preliminary injunction under § 10(j) of the NLRA. The district court granted that request and ordered Trinity to resume bargaining with the union. Although the Director is likely to succeed on the merits, she has failed to demonstrate that irreparable harm will result without an injunction. We therefore reverse the district court and vacate its order granting the injunction.

Trinity Health Grand Haven Hospital is a small community hospital serving patients in western Michigan. In years past, the facility operated under a different name. At the time, the hospital maintained affable relations with its employee union, which represented the hospital’s nurses, technical staff, housekeeping, and similar employees. But that relative peace soon ended when Trinity Health, a nationwide health system, acquired the hospital in June 2022.

A few months before Trinity took over, the collective bargaining agreement in place between the hospital and the union expired. To keep in step with its new negotiating partner, the union voted to affiliate with the SEIU, an international service workers' union. Trinity recognized the change of affiliation, agreed to extend the collective bargaining agreement through February 2023, and committed to negotiate a new agreement with the union down the road.

By April 2023, however, negotiations had yet to yield a definitive agreement. Around the same time, Jamie Quinn, a Trinity employee, approached her supervisor about getting rid of the union. After consulting NLRB resources, Quinn spearheaded an effort to collect her coworkers’ signatures in support of a petition to hold a vote on whether to remove the union, known as a “decertification” petition. With the help of a few sympathetic employees, Quinn gathered the Page 3requisite number of petition signatures. And following Quinn's filing of the petition with the Board in July 2023, the Board concluded there was sufficient support to call for a formal decertification election.

Trinity and the union entered into a Board -approved agreement to hold the election on September 18 and 19, 2023, with results to be announced soon thereafter. Yet the path to learning the election's results had its share of detours. As the parties were ironing out the election procedures, the union filed a request to block the counting of the ballots due to outstanding unfair labor practice charges it had previously filed against Trinity. This development, however, was not announced to Trinity. In fact, the Board did not inform Trinity or Quinn of the union’s request until the last day of voting. Nonetheless, at the election’s close, the Board impounded the ballots (as the union had requested) and did not reveal the results. A week later, on September 25, 2023, the union withdrew its blocking request. The Board in turn informed the parties that it would announce the results of the election on September 29.

The day before that announcement was to take place, Quinn delivered a so -called 'disaffection petition' (i.e., a petition to remove the union without an election) to Shelleye Yaklin, the hospital’s president. According to Quinn, she and her allies had continued collecting petition signatures after the election, supposedly from employees upset by the union’s blocking charge. All told, Quinn’s disaffection petition contained 94 signatures across seven pages. Yet it had several defects. Three of the seven pages—accounting for 45 signatures—did not display the petition’s statement of purpose, specifically, “We no longer wish to be represented by SEIU.” R.56 -1, PageID 3068. Nor were any of the petition signatures dated. In fact, 60 signatures had been recycled from the earlier decertification petition. Upon receiving the disaffection petition, Yaklin did little to verify the signatures included therein. At most, it appears Yaklin and Trinity’s human resources department tallied the raw numbers and determined that the total constituted a majority of Trinity’s roughly 180 -member bargaining unit. A few hours after Quinn delivered the petition, Trinity announced that it had received “objective evidence” of the union’s loss of majority support, meaning that Trinity would withdraw recognition of the union and that collective bargaining would cease. Id. , PageID 3068–69.

Page 4Trinity's declaration of victory proved premature. The Board announced the election results the next day. Of the 182 eligible voters, 89 cast ballots in favor of the union, while only 66 voted for decertification. Trinity objected to the election outcome, pointing to the disaffection petition, but the Board overruled Trinity’s objections. The Board then certified the union as the sole bargaining representative at the Grand Haven hospital. Nonetheless, Trinity dug in its heels and refused to recognize the union.

The union responded by filing unfair labor practice charges with the Board. In the union's view, Trinity's unilateral withdrawal of recognition and subsequent refusal to bargain violated the National Labor Relations Act. After an investigation, the Board’s Regional Director, Elizabeth Kerwin, issued a formal complaint against Trinity. The matter was tried before an ALJ, who ruled in the Director's favor. Emphasizing the disaffection petition's lack of credibility, the ALJ found that Trinity did not have “objective evidence” that the union had lost majority support when it pulled recognition. As such, the ALJ concluded that Trinity violated § 8(a)(1) and 8(a)(5) of the NLRA by refusing to bargain with a recognized union representative. (Trinity appealed the ALJ's ruling to the Board, but, because the Board has only recently regained a quorum, the matter has yet to be decided. See S. Res. 532, 119th Cong. (2025) (confirming two Board members); New Process Steel, L.P . v. NLRB , 560 U.S. 674, 683-84 (2010) (requiring at least three Board members to adjudicate cases).)

During the ALJ proceedings, the Director also petitioned the district court for a preliminary injunction under § 10(j) of the NLRA. On that front, the district court, following the ALJ's ruling on the unfair labor practice charged, concluded that the Director had made a sufficient showing to obtain interim relief. , No. 1:24 -cv -445, 2024 WL 4579090 (W.D. Mich. Oct. 25, 2024). The court in turn ordered Trinity to recognize the union and resume bargaining pending its Board appeal. Trinity has thus far abided by the injunction, and negotiations with the union remain ongoing.

At the same time, Trinity contests with us the district court's award of the injunction. After filing its appeal, Trinity promptly moved this Court for an emergency stay of the injunction during the pendency of its appeal, which we declined to order. Before us now is Trinity's principal request to dissolve the injunction.

Page 5On appeal, we ask whether the district court abused its discretion in granting a preliminary injunction, reviewing that court’s legal analysis de novo and its findings of fact for clear error. EOG Res., Inc. v. Lucky Land Mgmt., LLC , 134 F.4th 868, 874 (6th Cir. 2025). How does a district court abuse its discretion in this setting? '[B]y identifying the wrong legal standard, improperly applying the right one, or relying on clearly erroneous factual findings.” Id.

At issue here is the proper application of § 10(j) of the NLRA. That provision empowers the Board, upon issuing an 'unfair labor practice' complaint, to petition a district court for “appropriate temporary relief,” which the district court may grant “as it deems just and proper.” 29 U.S.C. § 160(j). Historically, our Circuit had read that statutory language as setting a low bar for the Board to clear to obtain a preliminary injunction; we would grant the Board’s request so long as there was “reasonable cause to believe that unfair labor practices ha[d] occurred,” and that relief was “just and proper.” McKinney v. Ozburn-Hessey Logistics, LLC , 875 F.3d 333, 339 (6th Cir. 2017) (citation modified). Not long ago, the Supreme Court corrected this notable misunderstanding of § 10(j). In Starbucks Corp. v. McKinney , 144 S. Ct. 1570 (2024), the Supreme Court rejected our Circuit’s “watered -down approach to equity,” id. at 1578, clarifying that before the Board may be awarded a § 10(j) injunction, it must satisfy the four-factor test courts traditionally apply when a litigant seeks a preliminary injunction, see id. at 1576. That is, the Board must make a “clear showing” that (1) it is “likely to succeed on the merits,” (2) it “is likely to suffer an irreparable injury in the absence of preliminary relief,' (3) 'the balance of equities tips in [its] favor,” and (4) “an injunction is in the public interest.” Id. (quoting Winter v. Nat. Res. Def. Council, Inc. , 555 U.S. 7, 20 (2008)).

Today is our first occasion to reconstruct our § 10(j) case law postStarbucks (which, we recognize, left the district court to go it alone in this instance). The wealth of other preliminary injunction precedents in our circuit, however, means we need not start from scratch. We know, for instance, that a preliminary injunction is an 'extraordinary and drastic remedy' that the Board should not get on easy terms. James B. Oswald Co. v. Neate , 98 F.4th 666, 672 (6th Cir. 2024) (quoting Munaf v. Geren , 553 U.S. 674, 689 – 90 (2008)). We also know that the second factor -irreparable harm —“is the core of the preliminary injunction,” without which the Board’s Page 6request for relief falls flat. EOG Res. , 134 F.4th at 883. To clear that high hurdle, the Board must show that the threatened irreparable harm is “both certain and immediate, not speculative and theoretical.' Memphis A. Philip Randolph Inst. v. Hargett , 978 F.3d 378, 391 (6th Cir. 2020) (citation modified). And we also know that although “irreparable harm” and “likelihood of success” often predominate, Nken v. Holder , 556 U.S. 418, 434 (2009), failure to make a clear showing as to any of the four factors sinks the Board’s request, regardless of how strongly other factors might favor the Board, see EOG Res. , 134 F.4th at 885 (citing Winter , 555 U.S. at 20). With all of this in mind, we now ask whether the Director has accomplished that tall order.

Likelihood of success on the merits. Start with whether the Director is likely to succeed on the merits of her unfair labor practice claims. Starbucks , 144 S. Ct. at 1576. Difficult though it may be to believe, our previous 'reasonable cause' test for assessing whether to issue injunctive relief never asked whether the Board (here, through the Director) in fact was likely to succeed in upholding its labor law determination. See McKinney v. Starbucks Corp. , 77 F.4th 391, 409 (6th Cir. 2023) (Readler, J., concurring), vacated , 144 S. Ct. 1570 (2024). That leaves us to first define what “success” means in this context. Critical to that consideration is defining before whom that success must be achieved. 'Congress,' it bears noting, 'did not make the Board’s remedial orders self -executing.” NLRB v. Starbucks , 159 F.4th 455, 471 (6th Cir. 2025). Instead, for the Board's order to have any real -world effect (i.e., success), a court of appeals must enter a decree enforcing that order. See id. (citing 29 U.S.C. § 160(e)). And for the Board to obtain that decree, the court of appeals must agree with the Board’s conclusion, reviewing the Board’s legal conclusions de novo and any fact findings for “substantial evidence.” Id. In view of these realities, we understand “likelihood of success” for § 10(j) purposes to refer not only to the Director’s chances of victory before the Board, but also the odds that this Court would grant a petition to enforce the Board’s order. See Frankl v. HTH Corp. , 650 F.3d 1334, 1355 (9th Cir. 2011).

That takes us to the merits issues upon which the Director must prevail. The Director charges Trinity with violating § 8(a)(1) and 8(a)(5) of the NLRA by unilaterally withdrawing its recognition of the union. Section 8(a)(1) prohibits employers from 'interfer[ing] with, restrain[ing], or coerc[ing] employees in the exercise of the right[]' to participate in union Page 7activity. 29 U.S.C. § 158(a)(1); see id. § 157. Section 8(a)(5), meanwhile, prohibits employers from 'refus[ing] to bargain collectively with the representatives of [their] employees.' Id. § 158(a)(5); see also id. § 159(a) (“Representatives [are] designated or selected for the purposes of collective bargaining by the majority of employees . . . .”). We have previously said that an employer violates these provisions by 'unilaterally withdrawing recognition from a union supported by a majority of the bargaining unit’s members.” NLRB v. Galicks, Inc. , 671 F.3d 602, 610 (6th Cir. 2012) (citation modified).

Here, the parties agree that the union, as the incumbent representative, presumptively enjoyed majority support at the time Trinity unilaterally withdrew recognition. See Auciello Iron Works, Inc. v. NLRB , 517 U.S. 781, 785 – 86 (1996). That leads the parties to assume that Trinity bears the burden to prove by a preponderance of the evidence that the union “had in fact lost the support of a majority of the unit employees.” Pac. Coast Supply, LLC v. NLRB , 801 F.3d 321, 324 – 25 (D.C. Cir. 2015) (deriving this burden-shifting framework from Board case law). It is not at all clear, however, that this Board-crafted presumption finds support in the NLRA itself or otherwise deserves any deference. See Loper Bright Enters. v. Raimondo , 144 S. Ct. 2244, 2269 (2024) (requiring “courts [to] exercise independent judgment in construing statutes administered by agencies”). That said, it makes no difference here who carries the burden of proof, so we take the case as presented to us. See United States v. Sineneng-Smith , 140 S. Ct. 1575, 1579 (2020). As a result, unless Trinity can show it had “objective evidence” that a majority of employees in the bargaining unit had abandoned the union, the Director will prevail on the merits. Wyman Gordon Pa., LLC v. NLRB , 836 F. App’x 1, 9– 10 (D.C. Cir. 2020).

Against this backdrop, the sole merits issue for the 'likelihood of success' factor is whether the disaffection petition supplied Trinity with sufficient evidence to justify its refusal to bargain. The Director contends the answer is a likely “no,” for two reasons. We find only one of these grounds persuasive.

1. 1. The Director’s chief argument is premised on the Board’s own cases holding that a properly conducted election trumps as a matter of law any other evidence of repudiation that might arise post-election. Appellee Br. 23 – 26; see, e.g. , Macy’s Inc. , 361 NLRB 1490, 1490 (2015) (“It is well settled that an alleged postelection loss of majority support is not relevant toPage 8 the question of whether a union should be certified as the result of a properly conducted Board election.”); Cmty. Support Network , 363 NLRB 833, 833 – 34 (2016) (applying this rule to decertification elections). Because Trinity does not contest the decertification election's legitimacy, the Director argues, the disaffection petition categorically fails given the election conducted less than two weeks earlier.

Measured by that standard, the Director’s success would seem all but certain. Yet her argument depends on a fatal oversight. By focusing on how the Board would rule based on its own case law, the Director takes too narrow a view of “likelihood of success.” Recall that the Director must make a 'clear showing' that she is likely to succeed before this Court as we engage in de novo review of the Board’s legal conclusions. And, it bears noting, “[w]e do not defer to the [Board’s] interpretation of the NLRA,” but instead use our “independent judgment.” Rieth-Riley Constr. Co., Inc. v. NLRB , 114 F.4th 519, 528 (6th Cir. 2024) (citing Loper Bright , 144 S. Ct. at 2262). That raises the question: Would this Court likely adopt the Board's categorical rule in favor of elections?

We think not. Start with the NLRA’s text. Trinity is first charged under § 8(a)(1) with 'interfer[ing] . . . in the exercise of the rights guaranteed' in § 7 of the Act. 29 U.S.C. § 158(a)(1). Section 7, however, does no more than announce the right to participate in union activity. See id. § 157. Trinity is also charged under § 8(a)(5) with 'refus[ing] to bargain collectively with the representatives of his employees, subject to the provisions of [§ 9(a)].” Id. § 158(a)(5). Yet § 9(a) simply defines the “representatives” with whom Trinity must bargain as those 'designated or selected for the purpose of collective bargaining by the majority of the employees' in the bargaining unit. Id. § 159(a). Notably, this language does not expressly require that those representatives be installed or removed by election, as the Board itself acknowledges. See Your Right to Form a Union , NLRB, https://www.nlrb.gov/about-nlrb/rightswe-protect/the-law/employees/your-right-to-form-aunion ('An election is not the only way a union can become your representative.'). And more to the point, nothing in the text of the statute provides that the results of a decertification election always trump any contrary evidence of union support.

Page 9True, § 9(e)(2) does instruct that no decertification election may be held in the “twelve -month period' following an earlier 'valid election.' Id. § 159(e)(2). For that reason, it may seem incongruous for the Act to ban decertification elections in the aftermath of an earlier election yet still allow for less formal challenges. Then again, the NLRA has long been understood to permit decertification by methods other than election. See, e.g. , Levitz Furniture Co. , 333 NLRB No. 105, at *1 (2001) (citing cases). So the Act’s limitation on decertification elections while simultaneously allowing for other avenues for decertification might reflect Congress’s understanding that elections are not the final word on union recognition. See City & County of San Francisco v. EPA , 145 S. Ct. 704, 714 (2025) (recognizing Congress’s omission of a statutory provision reflects on its intent). At the very least, this statutory language does not, as a matter of plain meaning, exclude the disaffection petition presented in this case.

The Director, for her part, offers no textbased argument to support the Board's interpretation. See Argument Tr. 29:00 – 30:00 (citing one of the NLRA’s purposes in promoting “industrial harmony”). Instead, the Board’s preference for elections seemingly rests entirely on a New Deal-era policy choice in favor of administrability. Id. ; see also Century Oxford Mfg. Corp. , 47 NLRB 835, 844 –47 (1943) (justifying the Board’s rule based on “the advantages of stability in collective bargaining' and ensuring that 'administrative determinations [do not] become ephemeral'). But whatever merit the Board's interpretation holds 'from a policy standpoint, it has no basis whatsoever in the statute’s text.” Lawson v. FMR LLC , 571 U.S. 429, 461 (2014) (Scalia, J., concurring).

Only the Supreme Court’s decision in Brooks v. NLRB , 348 U.S. 96 (1954), lends the Director some margin of support. There, employees voted to certify a union as their bargaining representative only to later present their employer with a signed letter indicating that the union had lost majority support. Id. at 97. The Board demanded that the employer recognize the union anyway, citing its one-year-certification rule, which required employers to honor elections for at least one year notwithstanding informal repudiation. Id. at 99. The Supreme Court sided with the Board. Critical to that outcome was the fact that Congress had “devised a formal mode for selection and rejection of bargaining agents” and “fixed the spacing of elections[] with a view of Page 10furthering industrial stability.” Id. at 103. As a result, the employees' 'informal repudiation' had no effect in the wake of a formal election. Id.

In some respects, that reasoning analogizes here. Yet the Board’s reliance on Brooks is ultimately unavailing. One reason is that Brooks spoke to the preclusive effect of certification elections that successfully install a new union, not decertification elections that, as here, do not result in a change of representative. Id. at 97. Only the Board itself, it seems, has extended Brooks to decertification elections. See Cmty. Support Network , 363 NLRB at 833 – 34. A second reason, one arguably more important than the first, is that Brooks upheld the Board’s pro -election rule largely in deference to 'the Board's discretion in carrying out congressional policy.' 348 U.S. at 104. Today, of course, we do not give the Board the same leeway in interpreting the NLRA. See Rieth-Riley Constr. Co. , 114 F.4th at 528 (“We do not defer to the NLRB’s interpretation of the NLRA.” (citing Loper Bright , 144 S. Ct. at 2262)).

All said, the Board’s interpretation has at best shaky legal grounding. It follows that the Director has not clearly shown how she will persuade this Court to disregard the disaffection petition as a matter of law. We thus disagree that the disaffection petition “misse[d] the mark legally.” Trinity Health , 2024 WL 4579090, at *8.

1. 2. Now turn to the Director’s backup argument. Even if Trinity were not categorically precluded from relying on the disaffection petition because of the prior decertification election, the Director explains, the disaffection petition nonetheless failed as a factual matter to prove the union’s loss of majority support. Recall that the disaffection petition contained 94 signatures, a slight majority of what Trinity understood to be the hospital’s 180 -member bargaining unit. That said, of the approximately 182 eligible voters (and of the 155 valid ballots cast) for the decertification vote, just 66 voted to decertify the union in the election (with 89 voting to keep it). Although the pro-union vote fell just shy of an absolute majority, it is difficult to reconcile the relatively low anti-union election turnout with the disaffection petition presented less than two weeks later. To the Director's mind, there is no credible explanation for this apparently sudden drop in union support. Along those same lines, the ALJ found that the disaffection petition contained so 'many defects' that it never gave 'objective evidence' of employees' wishes in the first place. Appellee Br. 28.

Page 11The Director puts forward meaningful evidence to support her point. First and foremost, of the petition's 94 signatures, only 34 could have been gathered after the election. The remaining 60 signatures had been submitted to the Director as part of the July 31, 2023, petition to hold the decertification election. That means some signatures had been gathered as early as May 2023 — a full four months before the election — thereby undermining their value. Confirming as much, the ALJ identified at least one employee who hurriedly signed the petition to hold the decertification election yet ultimately voted for the union. The absence of dates on the petition only compounds this issue.

What is more, three of the signed pages (accounting for 46 signatures) contain no language explaining the petition’s purpose. Although we decline to categorically disregard these pages as the Board has done in its own cases, see, e.g. , Wyman Gordon Pa., LLC , 368 NLRB No. 150, at *10 (2019), this reality further detracts from these signatures’ authenticity. For example, one employee who never saw the purpose statement unwittingly signed the petition believing it was about “better pay.” R.56 -1, PageID 3067 – 68. And, for one of the pages that did contain a purpose statement, its ten signatures lack a printed name or job title, leaving it difficult to readily identify the signers as union members.

Yet there are even more indicators that undermine Trinity’s disaffection determination. Within hours of Yaklin receiving the disaffection petition, Trinity declared it had received 'objective evidence' that the union had lost majority support. Id. , PageID 3068. Given that timing, taken alongside Yaklin's testimony, the ALJ's finding that Yaklin and others had not verified the petition’s accuracy will likely find support in “substantial evidence.” Starbucks , 159 F.4th at 462.

Trinity’s explanation of these events fails to convince us otherwise. Trinity first tries to square the disaffection petition with the election results. Here, Trinity highlights Quinn's testimony at the ALJ hearing that the union’s blocking of the election results caused “confusion” and “irritation” among Trinity’s employees. According to Quinn, employees continued to sign the disaffection petition after the election in opposition to the union’s meddling. Though perhaps a colorable explanation, it bears noting that the ALJ did not credit it because Quinn only cited the blocking charge as a galvanizing factor in response to a leading question. As for the disaffection Page 12petition itself, Trinity acknowledges its defects, arguing instead that other evidence authenticates the attached signatures. Namely, Trinity highlights the testimony from canvassers who said they showed signers the disaffection petition's purpose statement or otherwise orally explained its objective. At the same time, other testimony revealed instances of canvassers misinforming signers of the petition’s purpose. Taking all of this together, we see no basis to undermine the ALJ’s apparent conclusion that the petition process remained suspect.

Trinity’s arguments, we acknowledge, cast at least some doubt on the Director’s position. But two practical considerations push the Director's odds of success over the likelihood line. One, the disaffection petition garnered only a slight majority of Trinity’s bargaining unit. To lose even a few signatures would render the disaffection petition insufficient beyond dispute. Two, it bears remembering that this Court will eventually review the Board’s findings of fact on a deferential 'substantial evidence' standard. Starbucks , 159 F.4th at 462 (citing 29 U.S.C. § 160(e)). Should the Board adopt the ALJ's findings-for example, the ALJ's credibility determinations as to Quinn — that development would favor the Board in proceedings in this Court. See id. (citing Galicks , 671 F.3d at 607). Bearing this in mind, we ultimately agree with the district court that the Director has made a sufficiently “clear showing” that she is “likely to succeed on the merits.” Starbucks , 144 S. Ct. at 1576.

Irreparable harm. Even with the Director having established a likelihood of success on the merits, we will not grant her an injunction unless she also demonstrates that the Board is likely to suffer “irreparable harm” in the absence of injunctive relief. See Hargett , 978 F.3d at 391 (“Irreparable harm is an indispensable requirement for a preliminary injunction.” (citation modified)). When it comes to § 10(j) injunctions, irreparable harm means “harm that the Board, entrusted with its own enforcement powers, would otherwise be powerless to fix.” Starbucks , 77 F.4th at 409 (Readler, J., concurring) (citing Henderson ex rel. NLRB v. Bluefield Hosp. Co. , 902 F.3d 432, 440 (4th Cir. 2018)). On that front, it bears emphasizing that the NLRA empowers the Board to 'take . . . affirmative action' to remedy unfair labor practices. 29 U.S.C. § 160(c). That broad grant of equitable power authorizes an array of remedies including back pay, reinstatement, and, importantly here, an order to bargain in good faith. See, e.g. , FirstEnergy, LLC v. NLRB , 929 F.3d 321, 328 (6th Cir. 2019) (ordering bargaining and reinstatement of Page 13benefits). But see Starbucks , 159 F.4th at 478 (holding the Board lacks power to order consequential damages). Said differently, the Board’s remedial authority is far from “fragile.” Dissenting Op. at 24 . Against this remedial backdrop, it is perhaps no surprise that only in “rare situations' is interim relief needed to protect 'the Board's ability to remedy . . . unfair labor practices.” Sharp v. Parents in Cmty. Action, Inc. , 172 F.3d 1034, 1039 (8th Cir. 1999).

1. Today's case is not one of those unusual instances. Section 10(j) relief, we emphasize, “is an extraordinary equitable remedy that is never awarded as of right.” Starbucks , 144 S. Ct. at 1576 (citation modified). And the kinds of situations that justify such a “drastic” measure do not arise on a day-to-day basis. Munaf , 553 U.S. at 689 – 90.

Nonetheless, the Director believes she has cleared this high bar. Absent an injunction, the Director insists, Trinity’s refusal to recognize the union will erode union support to the point that the union will be unable to bargain effectively in the future. And that diminution in bargaining power, the Director’s reasoning goes, eludes the Board’s ability to fix going forward. According to the Director, this chain of causation is so 'obvious' that, in her view, absent “unusual circumstances,” courts can infer irreparable harm in refusal to bargain cases like this one, even where the Director lacks notable evidence demonstrating as much. Appellee Br. 37 – 38.

From where does the Director derive her argument? Primarily, decisions from the Ninth Circuit. Those cases hold that a court may (but need not) infer that an employer’s “refusal to bargain with a union” will cause irreparable harm due to the violation’s inherent “nature,” even if the Director offers no 'independent factual support.' Hooks ex rel. NLRB v. Nexstar Broad., Inc. , 54 F.4th 1101, 1115, 1118 (9th Cir. 2022) (quoting Frankl , 650 F.3d at 1362). In the absence of direction from our Court, the district court seemed to follow the Ninth Circuit’s lead, emphasizing that alleged labor violations similar to Trinity's have 'long been understood as likely causing irreparable injury to union representation.” Trinity Health , 2024 WL 4579090, at *9 (quoting Frankl , 650 F.3d at 1362). The dissenting opinion, too, insists we follow this same course. See Dissenting Op. at 25 – 27.

Page 14In the wake of Starbucks , we decline to adopt such a rule. There, the Supreme Court clarified that the Director must meet her burden the same way as would any other plaintiff seeking to demonstrate irreparable harm to secure a preliminary injunction. See 144 S. Ct. at 1578 –79 (explaining that the Board must qualify for relief under “traditional equitable rules” to obtain an injunction under § 10(j)). That is, like everyone else, the Director must clearly demonstrate “certain and immediate” harm; “speculative or theoretical” explanations will not do. Hargett , 978 F.3d at 392 (quoting D.T. v. Sumner Cnty. Schs. , 942 F.3d 324, 326 – 27 (6th Cir. 2019)); see also Winter , 555 U.S. at 21 (“just a possibility” of irreparable harm is insufficient) . Granting such an extraordinary remedy based on nothing but 'speculative and categorical' statements about supposedly “inherent” harm conflicts with the rare nature in which we award such relief. Bluefield Hosp. , 902 F.3d at 440. Indeed, if the irreparable harm the Director proffers here is so characteristically likely to flow from Trinity’s actions, it prompts one to ask why the Director cannot make a “clear showing” using affirmative evidence?

To be sure, every court resolving a preliminary injunction motion must inevitably make reasonable predictions about future harm based on existing evidence. See Dissenting Op. at 33. And it may sometimes be the case that evidence proving a violation of law (including a labor law) will overlap with evidence demonstrating irreparable harm. See id. But these realities do not warrant special rules favoring the Board. Whether characterized as a 'mandatory presumption' of irreparable harm in favor of the Board, or, alternatively, a 'permissi[ve]' inference of such harm, id. at 28 – 29 , we see no basis for putting a finger on the Board’s side of the scale, see Monsanto Co. v. Geertson Seed Farms , 561 U.S. 139, 157 –58 (2010) (“[A] court’s perfunctory recognition that an injunction does not automatically issue' does not justify a 'thumb on the scales.' (citation modified)). In the end, whether permissive or mandatory, Trinity’s alleged refusal to bargain, standing alone, does not mean that “irreparable harm can be inferred ,' Bluefield Hosp. , 902 F.3d at 440, without the Director offering 'independent factual support” of that harm, Hooks , 54 F.4th at 1118.

Seeing things otherwise, the Director calls our attention to the Second Circuit's post -Starbucks decision in Poor v. Parking Systems Plus, Inc. , 162 F.4th 335 (2d Cir. 2025). There, the appeals court reasoned that Starbucks did not disturb the Second Circuit's previous Page 15statements about irreparable harm in the § 10(j) context. Id. at 351. And because that court’s preStarbucks test paralleled our own, id. at 345, the Director explains, we should follow suit and allow permissible inferences for demonstrating irreparable harm. See Ahearn v. Jackson Hosp. Corp. , 351 F.3d 226, 239 (6th Cir. 2003) (inferring a “chilling effect” under the “just and proper” standard). Respecting how the Second Circuit might assess its own jurisprudence, it remains the case that our Circuit has long recognized that our former 'just and proper' test posed a significantly lower hurdle than the standard irreparable harm analysis we employ today. See Starbucks , 77 F.4th at 409 (Readler, J., concurring); Kobell ex rel. NLRB v. United Paperworkers Int’l Union , 965 F.2d 1401, 1409 n.3 (6th Cir. 1992). And with the Supreme Court’s Starbucks decision now our guiding light, we decline to follow prior cases that would artificially lighten the Director’s evidentiary burden.

The dissenting opinion, for its part, suggests that our approach departs from this Court’s customary practice of allowing “permissive inferences” of irreparable harm. Dissenting Op. at 27 – 2 8. But in the dissenting opinion’s lone cited case, we found the district court’s “findings” of irreparable harm 'well -supported' by 'independent' evidence of such harm to the plaintiff's reputation stemming from the defendant’s conduct. ACT, Inc. v. Worldwide Interactive Network, Inc. , 46 F.4th 489, 503 – 04 (6th Cir. 2022). For this reason, we expressly declined to reach the issue of whether a general presumption of irreparable harm can be recognized in copyright cases. Id. at 503. At most, we merely inferred that the independent evidence of defendant’s “history” of harming the plaintiff, plus the defendant’s “stated desire” to take similar actions, supported a showing of likely irreparable harm. Id. at 503 – 04. But we are aware of no case in which we have inferred harm based only on the “nature” of the act itself— the practice the Director asks us to employ. Frankl , 650 F.3d at 1362. Nor would we be wise to do so. After all, doing so would violate “traditional equitable rules.” Starbucks , 144 S. Ct. at 1578 – 79; see also ACT, Inc. , 46 F.4th at 503 (citing Winter , 555 U.S. at 22).

Indeed, not even the dissenting opinion is willing to adopt the aggressive inference the Director requests (and we reject) today. See Appellee Br. 37 ('[E]stablishing a likelihood of success on an unlawful refusal to bargain . . . establishes likely irreparable harm . . . .”); see also Hooks , 54 F.4th at 1118 (“[T]he dissent agrees that ‘independent factual support is not required Page 16when the court finds evidence of certain unfair labor practices.’” (quoting Frankl , 650 F.3d at 1129)). In its place, the dissenting opinion appears to prefer a middle ground, one that would allow the district court to infer irreparable harm if 'the facts indicate that' Trinity's alleged refusal to bargain “ha[s] already depressed union support.” Dissenting Op. at 25. Although we reject any Board-tailored special inferences in this setting, in the end, it may be that the dissenting opinion's framing renders our disagreement partially semantic. Perhaps, as the dissenting opinion itself acknowledges, the true source of disagreement stems from how one views what little independent evidence the Director has brought to bear. See Dissenting Op. at 29.

So we turn to that evidence now. To our mind, had the district court had the benefit of our decision today, which properly frames the Director’s burden, we think it would have taken a different view of the record and concluded that the Director has not demonstrated irreparable harm. Start where the Director does — with the union meeting attendance data from ten meetings between August 2023 and March 2024. The Director highlights that attendance hovered in the 20s in the runup to the election, but began falling after Trinity's withdrawal of recognition, having just eight attendees in a March 2024 meeting, suggesting that Trinity’s actions caused union support to decline. But at least two points deeply undermine that conclusion. One, it is far from clear that meeting attendance, which never broke more than a sixth of the overall bargaining unit membership and depended upon hospital employees' conflicting schedules, provides a meaningful gauge of support in the first place. Two, even accepting the Director’s premise, Trinity just as ably reads its own inferences into these numbers, in many ways through a more precise examination. As Trinity notes, attendance fell to nine — its second lowest point — immediately after the election and rose back into the 20s immediately after Trinity’s withdrawal. Equivocal data like this plainly cannot support a clear showing of irreparable harm.

The Director’s other piece of affirmative evidence is a witness affidavit submitted by a Trinity employee who served as the union president. The affidavit chiefly asserts, in broad terms, that Trinity’s withdrawal has caused “frustration and fear” among union employees. R.2 -16, PageID 165. According to the affidavit, employees “feel[] like” Trinity is “treating [them] like atwill employees,” with the affiant’s “overall sense” being that employees are confused and Page 17frustrated, with some “consider[ing] looking for other employment.” Id. , PageID 165 – 66. But hazy statements like these do not sufficiently demonstrate irreparably declining union support. At best, the affidavit cites just three specific examples of employee dissatisfaction: one employee feared discussing union organizing, another left Trinity for a better-paying facility, and another began considering doing the same. Even if these three anecdotes could reliably evidence hospital-wide trends, the affidavit still fails to answer the key question, namely, whether union support will not just decline, but in fact decline to the extent that the union will be unable to negotiate effectively by the time the Board issues a final order. See Bluefield Hosp. , 902 F.3d at 440. And on that score, the Director essentially leaves us guessing. As a result, this evidence too fails to reflect a “clear showing” of irreparable harm. See id. (declining to find irreparable harm based only on employees “feel[ing] frustrated” by employer’s “failure to bargain in good faith”); see also Hargett , 978 F.3d at 391 (“To merit a preliminary injunction, an injury must be both certain and immediate, not speculative or theoretical.” (citation modified)).

Switching gears, the Director alternatively argues that, absent an injunction, Trinity's employees will be irreparably deprived of benefits obtained through collective bargaining. The Director's witness affidavit highlights Trinity's post -withdrawal implementation of several policy changes, from “how overtime pay is calculated” to “holiday premium pay” to “the time required to not get a lunch from 5.5 hours to a 6 hour shift.' R.2 -16, PageID 166. By all accounts, however, these developments, if in fact harmful to union members, seemingly could be rectified with “back pay” or other equitable remedies within the Board’s purview. See 29 U.S.C. § 160(c). Falling back, the Director contends that collective bargaining often yields nonmonetary benefits, the loss of which the Board cannot easily rectify, such as “safety and health conditions,” “job security,” and “the protection of a grievance -arbitration procedure.” Appellee Br. 40. Accepting that generalized contention, however, would mean that the Director would be entitled to an injunction anytime collective bargaining is disrupted in any way. Her argument thus proves too much. And even entertaining her position here, the Director points to no record evidence showing how Trinity’s employees will be deprived of similar benefits. To the contrary, Trinity has not operated under a collective bargaining agreement since February 2023. To date, negotiations have been slowgoing, and the Director gives us 'no reason to assume' that the Page 18parties will reach a new agreement before the Board issues a final order. Bluefield Hosp. , 902 F.3d at 441 (citation modified). So this theory of harm gains no traction either.

More fundamentally, the Director fails to explain how Trinity’s unilateral changes will harm employees in the first place. Trinity, for example, began implementing pay raises shortly after ditching the union. The Director rationalizes that development by suggesting that union employees generally get higher wages. But generalizations aside, it remains conceivable that employers like Trinity might offer a better or different combination of perks to forestall union intervention, leaving workers at least as well off as with a union. That reality only cements our decision to decline the Director’s invitation to infer irreparable harm in refusal to bargain cases.

In reaching a different conclusion, the dissenting opinion sees the resolution here as entirely tied up in the district court's findings of fact. Dissenting Op. at 29. True, we must accept the district court’s “factual findings” absent “clear error.” EOG Res. , 134 F.4th at 874. But the district court did not render specific findings as to the Director’s affirmative evidence. See Trinity Health , 2024 WL 4579090, at *9. Likely because the district court made a “permissible inference” in the Director’s favor, the court made only brief mention of the meeting attendance data and witness affidavit, deeming them consistent with that inference. Id. At any rate, whether the Director has made a “clear showing” of “likely” irreparable harm is ultimately a legal yardstick, not a purely factual finding. See Starbucks , 144 S. Ct. at 1576 (citing Winter , 555 U.S. at 20). Taking the Director’s evidence at face value, she has not met that heightened evidentiary threshold. Indeed, the dissenting opinion recognizes that “reasonable minds” could differ on much of the Director’s evidence. Dissenting Op. at 30. In short, because all agree that the record paints an unclear picture, the Director has not made a “clear showing” of irreparable harm. Starbucks , 144 S. Ct. at 1576.

To recap: For the Director to have demonstrated irreparable harm, she needed to (1) specifically identify the injury threatened by the alleged unfair labor practice, (2) establish that the injury is likely to result absent an injunction, and (3) explain why the Board’s § 10(c) remedial powers could not fix the injury after the fact. And she must do so through more than mere speculation. See Hargett , 978 F.3d at 391. Because the Director has not completed each Page 19leg of that legal triathlon, she has failed to make a 'clear showing' of 'likely . . . irreparable harm.” Starbucks , 144 S. Ct. at 1576.

2. Were there any doubt on that score, the Director’s delay in seeking an injunction seals the deal. “An unreasonable delay in filing for injunctive relief will weigh against a finding of irreparable harm.” York Risk Servs. Grp., Inc. v. Couture , 787 F. App’x 301, 308 (6th Cir. 2019) (citing cases); see Cheetah Miner USA, Inc. v. 19200 Glendale, LLC , No. 23-1410, 2023 WL 6601863, at *3 (6th Cir. Oct. 10, 2023) (Delay “undercuts the sense of urgency that ordinarily accompanies a motion for preliminary relief and suggests that there is, in fact, no irreparable injury.” (citation modified)). In some cases, even a few months’ delay in filing a § 10(j) petition can work against injunctive relief. See NLRB v. Hartman & Tyner, Inc. , 714 F.3d 1244, 1249 (11th Cir. 2013) (finding that a delay of 'more than four months' was 'further evidence' weighing against § 10(j) injunction); Boire v. Pilot Freight Carriers, Inc. , 515 F.2d 1185, 1193 (5th Cir. 1975) (emphasizing that '[t]he Board waited three months' before filing a § 10(j) petition).

How long was the Director’s delay? The parties here dispute the answer. Trinity says the Director delayed seven months, starting from when Trinity first withdrew recognition. The Director, meanwhile, says she waited only four months, starting from when the Board issued the complaint against Trinity. Both are right, albeit in different ways. Section 10(j) did not allow the Director to petition for an injunction until the “issuance of [the] complaint” charging Trinity with an “unfair labor practice.” 29 U.S.C. § 160(j). And the charging and investigative process that resulted in the complaint no doubt took time. See id. § 160(b). So we accept the Director’s four-month timeline. Be that as it may, Trinity represents that the Director informed the company as early as September 26, 2023, that she intended to seek an injunction. That the Director seems to have had designs on § 10(j) relief from the start makes her four-month delay harder to explain.

The district court, for its part, did not decide which length of delay to credit, deciding instead that neither “conclusively augur[ed] against injunctive relief.” Trinity Health , 2024 WL 4579090, at *9. Even if not dispositive, however, the Director’s delay undercuts the “sense of urgency” implied by the Director’s evidence and “suggests that there is, in fact, no irreparable Page 20injury.” Cheetah Miner USA, Inc. , 2023 WL 6601863, at *3. More than that, we also see the delay as 'further evidence' that temporary relief will not add significant 'marginal benefit' beyond “a final Board order.” Hartman & Tyner , 714 F.3d at 1252; see Boire , 515 F.2d at 1193 (treating delay as “some evidence” that the employer’s unfair labor practices “have already taken their toll'). Here, the Director's affirmative evidence of irreparable harm almost entirely concerns setbacks faced by the union before the petition was filed. And the Director's predictions for the future, as we have noted, are at best speculative. Taken together, these considerations indicate that most (if not all) of the damage to the union was inflicted in the months before the Director petitioned for an injunction. See Hartman & Tyner , 714 F.3d at 1252. So the Director's delay reinforces our conclusion that the § 10(j) petition flunks the demanding irreparable harm inquiry.

* * *

On the whole, the dissenting opinion faults our approach as setting too high a bar. See Dissenting Op. at 28. The Director, the dissenting opinion reasons, must act before any damage to the Board's remedial powers 'metastasizes.' Id. at 25 . And because that 'accelerated procedure . . . limits the available evidence,” “permissive inferences” are an “essential tool” for the Director to show that harm. Id. at 25, 28. But the difficulties to which the dissenting opinion alludes bedevil every party seeking a preliminary injunction, not just the Director. And Starbucks , at its core, held that the Board must carry the same burden as any other party. See 144 S. Ct. at 1578 – 79. Although Congress could have lightened that burden, the fact remains that it did not. See id. at 1577. All in all, the Director's failure to demonstrate the long -established elements of irreparable harm compels us to dissolve the injunction. See EOG Res. , 134 F.4th at 884.

Balance of equities and public interest. Though not necessary to today's outcome, it deserves mention that the remaining injunction factors play a minor role in this case. In assessing the balance of the equities, courts go beyond the harm faced by the moving party and ask whether an injunction would harm the non-moving party even more. See EOG Res. , 134 F.4th at 885 – 86 (citing Winter , 555 U.S. at 24). The district court did not find convincing Trinity’s claimed “parade of horribles” should an injunction issue, and we agree. Trinity , 2024Page 21 WL 4579090, at *10. But as we find the Director’s theory of harm similarly unconvincing, the balance comes out essentially neutral.

With respect to the public interest, the Director cites “industrial peace” and a desire to ensure that Trinity does not profit from its alleged wrongdoing as important policies. Appellee Br. 45 – 46. Trinity counters that Starbucks demands more than platitudes about labor justice. That appears to be true. See Starbucks , 144 S. Ct. at 1576 (requiring a “clear showing . . . that an injunction is in the public interest”). But even taking the Director’s points as given, we find that public interest does little work here. As the district court tacitly observed, the parties' merits arguments largely subsume the public interest analysis. See Trinity Health , 2024 WL 4579090, at *11. After all, if Trinity was right to honor the disaffection petition as a true expression of its employees' wishes, then Trinity's actions harm neither 'industrial peace' nor the Board's remedial authority. Just the opposite, in fact: Employees’ right to remove a union they do not want stands on equal footing with their right to keep a union they do want. See 29 U.S.C. § 157 (protecting “the right to refrain” from union activity).

* * * * *

We reverse the district court and vacate its order granting the injunction under 29 U.S.C. § 160(j).

Page 22BOGGS, Circuit Judge, dissenting. I agree that the Director has demonstrated a likelihood of success on the merits, but would go further and affirm the injunction because she has made the required showing on all of the Winter factors. I particularly disagree with the majority’s approach for assessing irreparable harm, which mischaracterizes permissive inferences as mandatory presumptions.

Starbucks made headlines in our court, but it was old news for many of our peers. The Fourth, Seventh, Eighth, and Ninth Circuits have long applied the traditional four-factor preliminary-injunction test to § 10(j) petitions. The Fourth and Ninth Circuits in particular offer compelling guidance for assessing irreparable harm in this setting. Those courts have rightly rejected categorical or irreputable presumptions that a likely labor-law violation necessarily creates a risk of irreparable harm. But they did not prohibit district courts from drawing permissive inferences about the likely harm to the Board’s remedial powers based on specific factual findings, entitled to deference, about the observed and likely effects of challenged labor practices. Indeed, the Ninth Circuit expressly approved that approach, as has our own court in other legal contexts. The majority's reasoning breaks with the well -reasoned approach of our peers, however, leading me to respectfully dissent.

I

To obtain a preliminary injunction, a plaintiff “must establish that he is likely to succeed on the merits, that he is likely to suffer irreparable harm in the absence of preliminary relief, that the balance of equities tips in his favor, and that an injunction is in the public interest.” Winter v. Nat. Res. Def. Council, Inc. , 555 U.S. 7, 20 (2008). Congress has authorized the Board to petition a district court for a preliminary injunction (known as a “§ 10(j) petition”) to enjoin an alleged unfairlabor practice pending the Board’s resolution of the unfair -labor-practice claims Page 23and ultimate judicial enforcement of the Board's order. 29 U.S.C. § 160(j). The Board's regional directors — Kerwin, in this case — typically file the petition.

At least four of our peer circuits have long reviewed § 10(j) petitions under the same four-factor test applicable to all preliminary-injunction requests. Muffley v. Spartan Mining Co. , 570 F.3d 534, 542 – 43 (4th Cir. 2009); Bloedorn v. Francisco Foods, Inc. , 276 F.3d 270, 286 (7th Cir. 2001); McKinney v. Southern Bakeries, LLC , 786 F.3d 1119, 1122 – 23 (8th Cir. 2015); Frankl v. HTH Corp. , 650 F.3d 1334, 1362 (9th Cir. 2011). Until recently, however, we deviated from that approach by applying a diluted two-factor test that asked merely whether ''there is reasonable cause to believe that unfair labor practices have occurred,' and whether injunctive relief is ‘just and proper.’” Starbucks Corp. v. McKinney , 602 U.S. 339, 344 (2024) (quoting McKinney v. Ozburn-Hessey Logistics, LLC , 875 F.3d 333, 339 (6th Cir. 2017)). Our reasonable-cause standard was wrong. Finding that nothing in § 160(j)'s text 'jettison[s] the normal equitable rules” of preliminary injunctions, the Supreme Court instructed us to “apply the Winter framework” to § 10(j) petitions instead. Id. at 347 – 48 (citation modified).

Starbucks overruled our § 10(j) precedents, but it did not leave a blank slate. Nothing in the Court’s opinion disturbed our application of Winter in other contexts, nor did it question how the Fourth, Seventh, Eighth, and Ninth Circuits have applied Winter to § 10(j) petitions. See Poor v. Parking Sys. Plus, Inc. , 162 F.4th 335, 351 (2d Cir. 2025) (noting, in applying Winter to a § 10(j) petition for the first time, that Starbucks “did not disturb our precedents’ reasoning as to what constitutes harm justifying a grant of equitable relief'). So we retain the benefit of extensive authority from our own court and our peers to review the preliminary injunction issued below.

Unquestionably, the Director must demonstrate a likelihood of irreparable harm no matter the strength of her likelihood of success on the merits. Kentucky v. Biden , 57 F.4th 545, 555 (6th Cir. 2023). As the majority observes, irreparable harm in the § 10(j) context refers to harm to the Board’s remedial powers. Maj. Op. at 1 2; Frankl , 650 F.3d at 1362 (quoting Small v. Operative Plasterers' and Cement Masons' Int'l Ass'n, Local 200 , 611 F.3d 483, 494 (9th Cir. 2010))Page 24 (“permit[ting an] alleged unfair labor practice to reach fruition and thereby render meaningless the Board’s remedial authority is irreparable harm” (alteration in original)). My disagreement centers on the propriety of the logical framework that the district court used to reach its irreparable-harm conclusion. The majority faults the district court for relying on a presumption that because Trinity likely committed an unfair-labor practice, irreparable harm would inevitably result. Maj. Op. at 13 – 15. But that is not how the district court reasoned. Rather, it found specific facts showing that Trinity’s actions had already diminished union support and altered the parties’ bargaining positions, and inferred from those facts that, absent preliminary relief, the union’s position would continue to weaken, such that a final order restoring union recognition would prove futile. These permissive inferences, unlike mandatory presumptions, comport with Winter .

The Board’s remedial powers are fragile. Its core remedy for an unlawful withdrawal of union recognition is an order directing the employer to engage in good-faith collective bargaining with the union. See Muffley , 570 F.3d at 544. But the effectiveness of that final remedy depends on the resilience of union support and the original bargaining positions of the union and employer. Frankl , 650 F.3d at 1362 -63. If a union is 'weakened in the interim' between the inception of an alleged unfair-labor practice and the resolution of unfair-laborpractice claims, 'it will be difficult to recreate the original status quo with the same relative position of the bargaining parties.” Id. at 1363. And the bargaining process may fail altogether if the employer suppresses union support to the point that the workforce is no longer “inclined to unionize.” Poor , 162 F.4th at 352 (citation modified); see also Frankl , 650 F.3d at 1362 – 63. So, when the Director demonstrates a likelihood that an employer’s unlawful conduct will diminish union support below a critical level, preliminary injunctive relief is appropriate “to preserve the relative positions of the parties” during litigation. Starbucks , 602 U.S. at 346 (quoting Univ. of Tex. v. Camenisch , 451 U.S. 390, 395 (1981)). The district court permissibly found that the Director made that showing here.

A district court’s task in recognizing when to preserve the status quo often places it in a Catch-22. On the one hand, as the majority emphasizes, the requirement that a plaintiff make a “clear showing” that he is likely to suffer irreparable harm, id. at 345, necessitates an injury that Page 25is 'certain and immediate, not speculative or theoretical,' so a plaintiff cannot move too soon, D.T. v. Sumner Cnty. Schls. , 942 F.3d 324, 327 (6th Cir. 2019) (citation modified). On the other hand, plaintiffs fearing irreparable harm cannot wait too long, otherwise they will forfeit their claim to equitable relief through delay or obtain an injunction too late to reverse the damage. Churchill Downs Tech. Initiatives Co. v. Mich. Gaming Control Bd. , 162 F.4th 631, 642 – 43 (6th Cir. 2025) (describing the “Goldilocks zone” in which a request for a preliminary injunction is neither too speculative nor too late); Frankl , 650 F.3d at 1362 – 63 (recognizing the futility of the Board’s remedial powers if the detrimental effect of an unfair -labor practice is allowed to reach fruition). A preliminary injunction holds no value if irreparable harm has already occurred — otherwise, '[w]hy preserve a worthless option?' Samuel L. Bray, The Purpose of the Preliminary Injunction , 78 Vand. L. Rev. 809, 825 (2025) (explaining that '[t]he preliminary injunction protects the possibility of the final injunction”). “Judges are not engaged in a zero -tolerance prevention of all costs to the plaintiff,” id. at 854, but a polyp that threatens the Board’s remedial powers can be addressed before it metastasizes. Complicating the district court’s task here is the reality that it will never know exactly when a mere likelihood of irreparable harm crosses the threshold into reality. Such harms are often immeasurable, and the accelerated procedure for preliminary injunctions limits the available evidence. Id. at 821 – 26.

These considerations have prompted courts to develop analytical frameworks to navigate this uncertainty while preserving the extraordinary and discretionary nature of the preliminary injunction. Circuit courts with a long history of reviewing § 10(j) petitions under Winter have rightly rejected mandatory, irrebuttable, or categorical presumptions that impute irreparable harm from the mere likelihood of an unfair-labor practice. They permit, however, district courts to infer that the Board will likely suffer irreparable harm provided that the facts indicate that the alleged unfair labor practices have already depressed union support and altered the parties' relative bargaining positions. That is precisely how the district court proceeded here.

A sequence of Ninth Circuit cases illustrates the distinction between (prohibited) mandatory presumptions and (acceptable) permissive inferences. In 2011, that court affirmed a § 10(j) injunction stemming from an employer's alleged unlawful withdrawal of union recognition and failure to bargain in good faith. Frankl , 650 F.3d at 1358 – 61. Explaining the Page 26relationship between the likelihood-of-success and irreparable-harm inquiries, the Ninth Circuit recognized that the 'same evidence and legal conclusions, along with permissible inferences regarding the likely interim and long-run impact of the unfair labor practices that were likely to be found, preclude the conclusion that the District Court abused its discretion in finding a likelihood of irreparable harm.' Id. at 1363. The district court could infer that an extended exclusion of the union from collective-bargaining negotiations would likely irreparably harm the Board's remedial powers by depressing union support and distorting the parties' bargaining positions. See id. at 1362 – 63.

A decade later, the court clarified Frankl after some district judges read it to permit a finding of irreparable harm whenever the Director demonstrated a likelihood of success on the merits, even if the district court found no factual evidence supporting a risk of irreparable harm. Hooks v. Nexstar Broadcasting, Inc. , 54 F.4th 1101, 1118 (9th Cir. 2022). As in Frankl and our case, Hooks involved allegations of an unlawful withdrawal of union recognition. Id. at 1108 – 09. The Director requested a § 10(j) injunction and attempted to demonstrate a likelihood of irreparable harm through a loss of union support. Id. at 1110. When the district court specifically asked 'for evidence that employee support for the union was dwindling,' the Director mustered just one piece of evidence: testimony from a union representative that prounion employees feared retaliation. Ibid. But the district court found this evidence unpersuasive, likely because even 'in the face of alleged anti -union intimidation efforts,' the union 'stated that it collected signatures' from a majority of the relevant employees 'for a petition supporting the union.” Id. at 1108. Nevertheless, despite the district court’s finding that the record 'contained no factual evidence of irreparable harm,' it felt bound to 'apply the inference stated in Frankl ” to find a likelihood of irreparable harm. Id. at 1110 – 11.

The Ninth Circuit vacated the injunction because the district court failed to differentiate “between an impermissible presumption of irreparable harm when the Regional Director is likely to succeed . . . and a ‘permissible inference[] regarding the likely effects of that violation.’” Id. at 1116 (quoting Frankl , 650 F.3d at 1363). District judges cannot rely on presumptions of irreparable harm merely because the Director will likely succeed in proving a labor-law violation, because injunctive relief 'must be evaluated on a case -by-case basis in accord with Page 27traditional equitable principles and without the aid of presumptions or a ‘thumb on the scale’ in favor of issuing such relief.” Id. at 1114 (quoting Monsanto Co. v. Geertson Seed Farms , 561 U.S. 139, 157 (2010)). But in contrast to a mandatory presumption, which inappropriately relieves the party requesting the injunction of the burden of proof, a 'permissive inference' merely permits, but never compels, judges to reach conclusions supported by proven “predicate facts.' Id. at 1116 (quoting Francis v. Franklin , 471 U.S. 307, 314 (1985)). A permissive inference neither places a thumb on the scale nor reassigns the burden of proof; it simply allows a judge to use normal reasoning skills and common sense. And the judge “may rely on the same evidence . . . used to demonstrate a likelihood of success on the merits because some evidence may have probative value for both inquiries.” Id. at 1117.

The Fourth Circuit has held similarly, barring the use of mandatory presumptions without prejudicing the use of permissive inferences. See Henderson v. Bluefield Hospital Co., LLC , 902 F.3d 432, 441 –42 (4th Cir. 2018). There, the district court denied the NLRB’s § 10(j) petition, concluding that the Director had failed to demonstrate a likelihood of irreparable harm because the evidence showed “consistent commitment” to the union and substantial union activity even while the employer refused to engage in collective bargaining. Id. at 437. On appeal, the Director claimed that the district court “erred in refusing to infer harm from the nature of the violations.' Id. at 438. As in the Ninth Circuit, the Fourth Circuit rejected a categorical presumption that would require a finding of irreparable harm whenever the Director demonstrates a likelihood of success on the merits. Id. at 441. But because the record revealed minimal, if any, diminution in union support, Henderson did not restrict district courts' discretion to draw permissive inferences based on different facts in different cases. See id. at 441 – 42.

Applying Winter in other settings, our court has also approved permissive inferences as opposed to mandatory presumptions. Affirming a preliminary injunction in a copyright case, we found no problem with a district court projecting that the plaintiff would likely suffer irreparable harm based on infringement that had already occurred. ACT, Inc. v. Worldwide Interactive Network, Inc. , 46 F.4th 489, 503 – 04 (6th Cir. 2022) (Bush, J.). The established fact that a copyright infringer had already used infringing materials to compete with the plaintiff Page 28demonstrated a likelihood of irreparable harm because it 'suggest[ed] that [the infringer] will continue harming [the plaintiff’s] reputation, diminishing the perceived value of [the plaintiff’s] intellectual property, and unfairly competing with [the plaintiff].” Id. at 503. Similarly, Trinity has demonstrated a “history” of harming the union and makes no secret about its “stated desire” to end its reluctant relationship with the union as soon as the injunction dissolves, which will further damage union support. Id. at 503 – 04; see Appellant Br. 44 – 45. The combination of past harm plus a credible threat of future harm supported injunctive relief in ACT , and it should here too, because the Director has provided independent evidence of unfair-labor practices and diminution to union support.

Collectively, these cases rightly rejected the use of mandatory presumptions that collapse the likelihood-of-success-on-the-merits and irreparable-harm inquiries. Left unchecked, mandatory presumptions would make preliminary injunctions in this context automatic rather than discretionary, frequent rather than extraordinary, and merits-focused rather than remedycentric. See Henderson , 902 F.3d at 440 – 41; Bray, supra , 827 – 34. But these cases also correctly allowed district courts to draw inferences about the future based on the facts describing the present. See ACT , 46 F.4th at 503 – 04; Hooks , 54 F.4th at 1116. Today, the majority rejects that approach for evaluating the Director's request for injunctive relief. This is a mistake: barring district courts from employing permissive inferences will needlessly cramp their discretion and deny them an essential tool to reconcile often-limited evidence in the preliminaryinjunction context. And it will undermine the Board’s remedial powers by allowing unfair -labor practices to break union support while litigation crawls ahead. The Board is certainly not entitled to special treatment, see Maj. Op. at 16, but neither should we disadvantage it relative to other litigants, see ACT , 46 F.4th at 503 – 04.

The district court followed this approach, applying the correct legal standard because it employed a permissive inference rather than a mandatory presumption. It expressly refrained from assuming irreparable harm from the Director’s likelihood of success on the merits, and it rightly assigned to her the burden of obtaining relief. R. 68 at 17 – 19, PageID 3244 – 46. True, the court recognized that some of the same evidence relevant to the likelihood-of-success determination, such as evidence of the hospital’s failure to bargain in good faith, could bear on Page 29the likelihood that union support would deteriorate substantially while litigation proceeds, risking irreparable harm to the Board’s remedial powers. But the district court also consulted evidence that it had not considered in its likelihood-of-success analysis, such as data indicating a decrease in attendance at union meetings after the hospital withdrew recognition and an affidavit describing specific examples of how the withdrawal of recognition chilled union activity. This evidence, not any abstract assumption, formed the basis of the district court's 'permissible inference.” R. 68 at 18, PageID 3245. Reasonable minds may debate the strength of the factual predicates for the district court's inferences, as I will address shortly, but the district court undoubtedly avoided the legal error of presuming a likelihood of irreparable harm despite the absence of any record evidence showing such a risk. Hooks , 54 F.4th at 1110 – 11, 1116 – 17.

As other circuits have approved in reviewing § 10(j) petitions, and applying reasoning we have approved in other settings, the district court found facts indicating that harm to the Board’s remedial powers had started to accumulate, and inferred that such harm would become irreparable without preliminary relief. I would continue to allow permissive inferences in the § 10(j) context to avoid creating conflict between the circuits and within our own court.

Perhaps the majority’s core rationale for reversal lies not in the district court’s usage of permissive inferences per se, but rather in the facts supporting its inferences. Clearly, the majority disagrees with the district court’s assessment of the facts indicating that the hospital’s withdrawal of recognition diminished the union’s support and bargaining position. Maj. Op. at 16 – 18 . But we owe great deference to the district court's factual findings. In conducting our “highly deferential” assessment of whether the district court abused its discretion in granting the preliminary injunction, we review its factual findings for clear error. Churchill Downs , 162 F.4th at 637 (quoting Certified Restoration Dry Cleaning Network, LLC v. Tenke Corp. , 511 F.3d 535, 541 (6th Cir. 2007)). “A finding is ‘clearly erroneous’ when, although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed.” Certified Restoration , 511 F.3d at 541 (quoting Anderson v. City of Bessemer City , 470 U.S. 564, 573 (1985)).

Page 30The district court did not clearly err in finding that Kerwin made an adequate showing of irreparable harm because the facts indicate that union support had meaningfully declined, permitting an inference of continued deterioration absent preliminary relief. In addition to citing the hospital’s failure to bargain in good faith with the union, the district court found a likelihood of irreparable harm based on a decline in attendance at union meetings and an affidavit from Ricky Kauffman, a union member who became the bargaining-unit president in November 2023, that described the effects of the hospital’s withdrawal of recognition. R. 68 at 18, PageID 3245. I agree with the majority to the extent that reasonable minds could draw different conclusions in the first instance from the union-attendance data and the affidavit. But if reasonable minds could differ, then the district court did not clearly err in assessing that union support had declined. See Certified Restoration , 511 F.3d at 541. The district court made reasonable findings from uncontroverted evidence, and additional facts in the record support its conclusions. Applying an appropriate legal standard to analyze these facts, the district court did not abuse its discretion in concluding that Kerwin clearly showed that the Board will likely suffer irreparable harm without this injunction.

Considering the evidence in turn, the district court did not clearly err in finding that the attendance data indicates a loss of union support. The record reports attendance at nine union meetings, conducted virtually, from August 2023 through March 2024. R. 2-15, PageID 162. Six of those meetings preceded the hospital’s withdrawal of recognition. Those meetings were attended by 23, 24, 30, 29, 16, and 9 union members (the last of these, held on September 21, 2023, occurred after the decertification election but before the hospital's withdrawal of recognition). The other three meetings followed the hospital’s withdrawal of recognition. On October 3, 2023, roughly a week after the withdrawal of recognition, 23 members attended. On October 25, 2023, only a month after the withdrawal of recognition, attendance dropped sharply to 11. The final meeting, held on March 4, 2024, had only 8 attendees.

Perhaps, if I were a district judge, I might find this data insufficient to conclude that union support had begun to waver. But as an appellate judge owing deference on matters of fact, I cannot conclude that the district court clearly erred, given that the post-withdrawal meetings had, on average, just two-thirds the attendance of the pre-withdrawal meetings (14 compared toPage 31 21.83) and that two of the three lowest-attended meetings occurred post-withdrawal. The majority, however, perceives a different trendline. It emphasizes the low attendance at the last meeting before withdrawal (9 members on September 21) and high attendance at the first meeting after withdrawal (23 members on October 3) to discount the Director’s evidence. Maj. Op. at 16 . But the majority’s second -guessing ignores those meetings’ contexts. The September 21 meeting occurred after voting concluded in the decertification election, but before the tally was revealed or the hospital withdrew recognition. Set amidst that holding pattern, there was probably not much business to conduct or reason to attend. In contrast, more union members understandably flocked to the October 3 meeting to reckon with the immediate implications of the hospital’s withdrawal of recognition. The majority attributes more weight to those two data points than their unique circumstances can bear. Far more instructive, in my mind, is the fact that attendance had precipitously diminished by October 25, after the election dust settled but still within a month of Trinity’s withdrawal of recognition, and never recovered.

Kauffman’s affidavit also supports the district court’s permissive inference. His affidavit describes how the hospital's withdrawal of recognition and refusal to bargain with the union chilled union participation. R. 2-16 at 2 – 3, PageID 164 – 65. He explains that a fellow bargainingcommittee member resigned her leadership position 'because she felt like it was futile to continue to fight [Trinity].” R. 2 -16 at 2, PageID 164. Other employees felt pressured to refrain from even discussing the union at work, especially as supervisors informed new hires 'that there was no union at the facility and that there never would be again,' suggesting a growing sentiment of futility and fear. R. 2-16 at 2 – 3, PageID 164 – 65. The affidavit also describes various policy revisions that Trinity instituted unilaterally after withdrawing recognition, including changing employees' healthcare benefits, cutting premium -pay benefits, revising overtime calculations, increasing the shift length required to receive a lunch break, and misclassifying employees to exclude certain bargaining-unit members from pay increases. R. 216 at 4, PageID 166. This account of the hospital’s sweeping changes are consistent with the ALJ’s finding that the hospital unilaterally rescinded 16 policies after withdrawing recognition, R. 561 at 33, PageID 3071, lending support to the district court’s decision to credit Kauffman’s affidavit.

Page 32The corroborated fact that the hospital altered many policies governing pay and conditions of employment reinforces the conclusion that the Director showed a likelihood of irreparable harm. As I have explained, the irreparable-harm inquiry in this context assesses the potential damage to the Board’s remedies, such as ordering the employer to resume good -faith collective bargaining, by considering support for the union and the original bargaining positions of the union and employer. See Muffley , 570 F.3d at 544; Frankl , 650 F.3d at 1362 – 63. That damage is not speculative here: “affirmative evidence,” Maj. Op. at 14, from multiple sources shows that the hospital's unilateral actions have significantly distorted the parties' bargaining positions. If the Board were to ultimately order a bargaining remedy, the union would need to recover the ground lost by the hospital's unilateral policy changes before it could bargain for additional benefits. By widening the gap between its own position and the union's in future rounds of bargaining, the hospital has weakened the Board’s remedial powers.

The Director timely acted upon these risks. She petitioned for injunctive relief within four months after she filed unfairlabor charges challenging Trinity's withdrawal of union recognition, the event that triggered her statutory authority to petition the district court. See 29 U.S.C. § 160(j). The majority disputes that starting point, but accepts that seeking injunctive relief within four months would be prompt enough. Maj. Op. at 19 (citing NLRB v. Hartman & Tyner, Inc. , 714 F.3d 1244, 1249 (11th Cir. 2013)). Even if the Director's clock started when Trinity withdrew recognition, a seven-month delay would still fall within the timeline that courts have accepted for granting a preliminary injunction. Muffley , 570 F.3d at 544 – 45 (affirming injunctive relief requested eighteen months after the NLRB filed its complaint). Either way, the Director timed her petition appropriately, requesting injunctive relief after harm had begun to accrue but before it became irreparable.

There is a different delay worthy of concern, however: the years-long wait for the Director to obtain final relief. The Board only recently regained a quorum, and it presumably must toil through a considerable backlog. Maj. Op. at 4; see Poor , 162 F.4th at 352. Even after it adjudicates Trinity’s administrative appeal, the Board would still need to petition for judicial enforcement of its order. Maj. Op. at 6. This litigation timeline exacerbates the risk that the initial decline in union support and distortion to the parties' ex ante bargaining positions will Page 33continue unabated, rendering worthless the NLRB’s final remedy of ordering the employer to recognize the union.

In sum, while the evidence could have led the district court to a different conclusion, it sufficiently supports its irreparable-harm finding to preclude us from substituting our judgment.

Briefly, I also disagree with the majority regarding the balance of the equities and the public interest. The balance of the equities favor the Director because Trinity's 'parade of horribles” fail to persuade, Maj. Op. at 20, so no cognizable harm to the employer outweighs the Director's irreparable -harm showing. The public interest also favors injunctive relief. The National Labor Relations Act, by its plain text, protects the unionization preference of the bargaining-unit majority. See 29 U.S.C. §§ 157, 158(a)(1). Because the evidence here indicates that a majority of the bargaining-unit employees supported the union, see Maj. Op. at 10 – 12, the public interest lies in vindicating that preference. See Frankl , 650 F.3d at 1365 (quoting Miller v. Cal. Pac. Med. Ctr. , 19 F.3d 449, 460 (9th Cir. 1994) (en banc)) (“In § 10(j) cases, the public interest is to ensure that an unfair labor practice will not succeed because the Board takes too long to investigate and adjudicate the charge”). The Director has satisfied the third and fourth Winter factors.

The majority opinion understandably seeks to preserve Winter ’s rigor, but it travels too far in the opposite direction. Any preliminary-injunction analysis must accommodate some predictive reasoning — the point, after all, is to prevent irreparable harm before it occurs. When reviewing requests for preliminary injunctions of all stripes, including § 10(j) petitions, our court and others have long permitted district courts to find a likelihood of irreparable harm if factual findings about the present support permissive inferences about the future. This framework does not convert the preliminary injunction from an extraordinary and discretionary remedy into an ordinary and automatic one. See Hooks , 54 F.4th at 1119 – 20 (vacating a preliminary injunction); Henderson , 902 F.3d at 440 – 43 (affirming the denial of a preliminary injunction). Rather than compelling pro-Board outcomes, it merely asks the right questions: are there facts clearly Page 34indicating a present weakening of the union’s support and bargaining position, and do those facts support an inference that the union’s strength will continue to decline before the Board can issue and enforce a final bargaining-order remedy? Because the majority opinion departs from this framework, creating inter- and intra-circuit conflict in the process, I respectfully dissent.