After #65206 fix, I disabled use_hedged_requests on production, that has 24.2.2.71 (without upgrading), but, surprisingly it did not help fully, I still saw crashes from time to time.
And I think that the culprit is async_query_sending_for_remote (since I saw sendQueryUnlocked in stacktraces)
I've read the code around, found couple of issues, but I still don't see how this use-after-free case for fibers is possible, I'm going to continue digging, but just to document this thing somewhere I've created this issue.
If anybody has seen such crashes (or maybe someone already saw some of them on CI), please attach full logs for queries/relevant threads here.
Cc: @Avogar
Refs: #47229