Termination Detection in an Asynchronous Distributed System with Crash-Recovery Failures

We revisit the problem of detecting the termination of a distributed application in an asynchronous message-passing model with crash-recovery failures and failure detectors. We derive a suitable definition of termination detection in this model but show that this definition is impossible to implement unless you have a failure detector which can predict the future. We subsequently weaken the problem and strengthen the failure model to allow solvability