Better reporting of cluster failure when running replicates in parallel

[pacematt]

Currently, errors generated by replicates result in the cluster failing and all replicates in that cluster are absent from the returned results object. This is not in-and-of-itself a bad system, but the user needs to be alerted to the loss of replicates.