Fallback server

[l4l]

If the coordinator fails then the whole network goes down. Would be really nice if it's possible to set-up a fallback server(s) which turn the control if the main one fails.

Ideally that shall work over some consensus algorithm (e.g raft) to make the proper sync and handle the case when certain peers cannot connect to the first server and other kinds of network partitions.

Might be related to #290

[strohel]

If the coordinator fails then the whole network goes down.

To be more precise: new clients cannot connect and (public IP) updates to existing ones won't propagate, but already established p2p connections remain even if the server stops.