> There's probably not much more we can learn at this point --- given
> that the entry is old, any other evidence is probably long gone.
I'm afraid I'll never be able to get soon enough to the evidence even if
it happens again, unless I manage to reproduce it myself, which I tried
but didn't succeed (100 threads competing for ~10 connections
creating/dropping the table 1000 times, with ~10 percent of the cases
rolling back before the drop).
> I'd suggest just doing a manual DELETE of that pg_type row to get out
> of the problem.
OK. This was not a huge problem, it's a test system, but enough ground
for the QA to scream ;-)
> And yes, you ought to update that 8.1.0 installation ;-). But I have
> no idea if this problem is related to any solved bugs.
The QA installed the right version now.
Thanks for the assistance.
Cheers,
Csaba.