Can anyone identify a likely cause of the stack traces reported here:
https://retrace.fedoraproject.org/faf/reports/853674/
Apparently, what's happening is that PGSemaphoreUnlock is hitting
elog(FATAL, "semop(id=%d) failed: %m", sema->semId);
and then that's being promoted to a PANIC, presumably because we're
inside a critical section.
I have no more data than what's on that webpage --- I just got an
auto-notification that the number of similar reports had reached 10.
Those might all be from the same Fedora user, for all I can tell;
in which case it might have some pedestrian explanation.
regards, tom lane