in-kernel PM: re-establish subflows after "network" errors

[matttbe]

In case of network issue (blackhole, wrong path, etc.), it might be interesting to re-established subflows that got closed after a TCP-level error (timeout, RST, etc.) as it was maybe a temp issue or a change in the network (e.g. NAT dropping connections a bit too quickly)