In the event the network becomes poorly connected, we can encourage convergence on a stake winner by accepting the highest ranking proof that we know about from remote proofs. The vast majority of the time, this will be the same proof selected as the local winner, and so will have no change in behavior.