Maybe mwlog1001 will have some details about what happened?

First of all, the edit was revdel'ed, hence I cannot help with on-wiki debugging in any way. Would it be possible to undelete it, at least temporarily?

That said, I've checked logstash to see if there's something interesting for that time. Lately, the AF channel is clogged by T230256, which I hope gets resolved soon. At least, I think that unset variables are not the cause of this bug, given that the filter only uses variables available for all edits, and the action in question was an edit. However, all log entries on logstash for that time (12/01, ~11:27 UTC) refer to unset variables.

We may try grepping logs on mwlog, but I don't think we'll find anything useful. Recently, I've reported T241519, which seems very similar, but we didn't find anything useful in the logs.

Given that this is not the first time this bug happens, I'm wondering whether we should add a debug line at the end of checkAllFilters, e.g.

$logger->debug( __METHOD__ . ' filters matched: {filters}', [ 'filters' => implode( ', ', $matchedFilters ) ?: 'none', 'user' => $this->user->getName(), 'title' => $this->title->getPrefixedText() ] );

to be executed unconditionally for all actions. That would certainly help us understand whether the issue is with parsing the filter or with taking consequences. Note, however, that it would add gazillions of log entries. To reduce the confusion, I'd like to wait for r538598 before moving on.