Implement better handling for various remote comms failure modes

[FUDCo]

The current remote comms implementation (#578) mostly implements just the happy path with respect to network communications. For this software to be practical, it needs to be much, much better at handling the various failure modes that can crop up with respect to peer-to-peer messaging, including dealing with both permanent and intermittent unavailability of peer nodes, failure of network communications pathways, failure of essential intermediaries, and no doubt a much larger set of things that we as yet haven't really contemplated very deeply.

We need to both do a more comprehensive analysis of the possible failure cases and then devise and implement mechanisms for coping with them. This will mostly likely unpack into a series of more specific issues as the various ways things can go off the rails get identified.