I think we've started to venture into system-level territory that the standard library itself shouldn't have to account for. If TCP on systems are configured by default to allow for infinite timeouts, then it should likely be an issue for those distros. I don't think accounting for such things should be the responsibility of the standard library. The socket module behaves appropriately in handing off timeouts to be managed by the system level if left undefined. I think that any further discussion around timeouts should be taken up on distro mailing lists, or even perhaps python-list as it's not directly relevant to this issue.

As for this specific issue, unless someone beats me to it, I'll get a patch together introducing the session_factory as soon as I can.