As always it depends on the circumstances, but should default to quietly closing with WONTFIX. Others have said Daniel is typically helpful and respectful so there we go.
What you linked to is not really evidence, just an unsubstantiated allegation. Over the top public shaming is something that should be pretty easy to provide direct evidence of. When Linus Torvalds does it, it gets repeatedly brought up in forums like this for many years.
> I have no reason to believe it is a lie, and it sounds plausible
Except for all of the responses from people saying it doesn't sound plausible for the project in question, and for the acute lack of real evidence or even details to accompany the allegation.
Additionally, I think "last resort" is way too high a bar. It's totally reasonable for an open source project to have a zero-tolerance policy for AI-generated spam patches or bug reports, and to respond with public shaming after the first offense. Nobody should be expected to make any allowance for such egregious behavior.
A user who is genuine but simply doesn't know how to usefully communicate about their problem doesn't deserve that treatment and should simply be ignored if the devs don't have time to engage in the interrogation necessary to extract a useful bug report. But if the user decides to try to use an LLM to compensate for a lack of content in their bug report, that user would be earning a negative response by making a bad decision. (If you're going to use an LLM, ask it how to write a bug report, rather than asking it to make up a bug report for you.)