but, yea, the client software is pretty stable
when it fails it just reconnects
which is what makes it stable :)
even to another node?
no, sadly not
i really should implement that
but then it could happen that it switches to a node with higher latency for a small error and does not switch back
yes
issues to consider :)