"Brendan O'Shea" <brendanoshea(at)comcast(dot)net> writes:
> Is there no way to specify a timeout for the write() to the socket or some
> other way to abort?
This is really a question to take up with your TCP stack implementors.
I think it is fundamentally wrong for Postgres to be second-guessing
the network software about whether a network connection is still live.
regards, tom lane