This article A Purpose-Built Global Network: Google’s Move to SDN makes reference to an interesting concept:fate-sharing. The article and wikipedia entry both use the term in the context of network protocol/architecture design but the concept has broader applicability.
The article does address the concept the context of a distributed system, referencing that problems that occur when a switch fails independently of its controller or vice-versa.
Consider Leslie Lamport’s famous quote “A distributed system is one in which the failure of a computer you didn’t even know existed can render your own computer unusable”.
A non distributed system has only one machine and everything fails when it fails. There is no need to for failure detection or attempting if a slow response means the remote system has failed (or is merely slow). It also means that a reboot might fix things and is unlikely to make matters worse.
The STONITH principle attempts to reduce the impact of the problem by forcing a machine to a known state.
The Akka cluster quarantines nodes to ensure they remain dead.