Tim Tassonis <timtas@cubic.ch> writes:
> I try to reproduce the situation tomorrow and will also check on any
> odbc_errmsg() messages and the postmaster log.
OK. A couple of comments: the only timeout within Postgres itself is
statement_timeout, which I think wouldn't apply to your situation even
if you had it set (but I might be wrong). We have also heard reports
of firewalls dropping idle connections, but in that scenario you'd
not still have a working connection afterwards.
regards, tom lane