Robert Creager <Robert_Creager@LogicalChaos.org> writes:
> What more information would you be looking for?
Well, in the words of the old saw, if we knew what we were looking for
it wouldn't be research ...
I'm sure we could solve it quickly if we could get a reproducible test
case, but it sounds like you are nowhere near being able to provide
that. I would suggest looking for context information: what else is
happening at the same time this happens?
Another thing that would probably give enough information to solve it
is a trace of the connection from a TCP packet sniffer, for at least
a few packets leading up to the error message. Not sure if it's
practical to try to get one, but if you could it'd be great.
regards, tom lane