You just assume network partition is never happening. If calls are synchronous, assume they always reach other side and finish. If async, also assume successful

You just assume network partition is never happening. If calls are synchronous, assume they always reach other side and finish. If async, also assume successful outcome immediately after initiating the call. This "almost always" works.