"Chris White (cjwhite)" <cjwhite@cisco.com> writes:
> Also, I found by re-indexing the pg_largeobject table the problem seems to
> have gone away.
Um. That's probably destroyed the evidence :-( ie gotten rid of
whatever weird state you'd gotten the index into. But if you have
any luck reproducing it, I definitely want to know.
regards, tom lane