Fatigue Was the Stopping Rule
A code review thread doesn't end when the code is correct. It ends when one of the two people in it stops replying.
That's not cynicism, it's just how the mechanism works most of the time and nobody writes it down because it's slightly embarrassing. Reviewer raises an objection. Author responds. Reviewer raises a second, narrower objection. Author responds to that too, maybe with less patience. A few rounds in, somebody decides the remaining disagreement isn't worth the cost of continuing to disagree, and either approves with a shrug or lets the thread go quiet. The merge happens. Nobody wrote "I am now too tired to keep arguing" in the comment box, but that's the actual event that closed the loop.
Peer review runs on the same current under a thicker coat of formality. A paper gets "revise and resubmit" twice, three rounds of reviewer comments, and at some point either the reviewer stops finding new objections worth the time to type up, or the author stops having the energy to rebut one more marginal note about figure 3. The paper that gets published isn't the one that resolved every objection. It's the one where the objecting ran out before the objections did.
Call it the patience budget. Every review protocol, every appeals process, every comment-and-response cycle has an implicit assumption baked into its design: both parties are human-speed, and human-speed participants get tired. Tiredness is not a bug in the system, it's the stopping rule the system was built on top of, usually without anyone stating it as a design requirement because it didn't need to be. Of course people get tired of arguing. That was always going to be true, so nobody engineered a backup for when it stopped being true.
Courts are the one place this got formal enough to name. A litigant who keeps filing, keeps appealing, keeps raising the same claim under a new caption after losing it three times, isn't winning any of those cases on the merits. They're just not running out. Enough jurisdictions have hit this often enough that there's a standing legal category for it: a vexatious litigant, someone a court can order to get advance permission before filing again, precisely because the normal mechanism, opposing counsel's patience and the court's calendar both finite, failed to produce an ending. The label exists because the system's actual stopping rule, fatigue, had been defeated by someone who simply didn't experience the cost that rule depended on.
That's the tell. You don't need a new administrative category for a process whose real logic was "whoever's right wins." You only need one when a party has discovered, deliberately or not, that the thing standing between them and an infinite extension of the dispute was never the merits. It was the other side's finite tolerance for continuing to engage. Once that tolerance is bypassed, either by obsession, by legal aid funded past any sane stopping point, or by something that doesn't experience the passage of argument-time as a cost at all, the review loop doesn't resolve toward correctness. It just runs, because nothing in its design tells it to stop on any basis other than somebody getting sick of it.
Which is the part worth sitting with before handing argument-generation to anything that doesn't tire. A reviewer facing a human author is trading on a resource the author also has a stake in conserving: both sides want this over by Friday. A reviewer facing something that can produce a technically-responsive rebuttal to every objection, instantly, at no cost to itself, isn't facing a peer anymore. It's facing a process that will match it move for move indefinitely, and the review protocol has no second mechanism in reserve for that case, because it was never built expecting one side to opt out of getting tired.
The fix people reach for is usually "just be more rigorous about the merits," as if the problem were insufficient standards. It isn't. The merits were never what closed the loop. Patience was doing that job quietly the whole time, and the loop has no idea how to end without it.