In the event of a server crash, a solution for recovery can be:
1) Proxy picks a server (at random?) to roll back changes
2) Server determines affected clients (i.e. clients who have in their view pieces who was in the responsibility of the crashed server).
3) Tell clients to stop (or move their frustrum somewhere else) and wait for rollback.
4) When rollback has been confirmed, release affected clients
5) If crashed server was coordinator, promote new leader election
In the event of a server crash, a solution for recovery can be: 1) Proxy picks a server (at random?) to roll back changes 2) Server determines affected clients (i.e. clients who have in their view pieces who was in the responsibility of the crashed server). 3) Tell clients to stop (or move their frustrum somewhere else) and wait for rollback. 4) When rollback has been confirmed, release affected clients 5) If crashed server was coordinator, promote new leader election