Parallel mode should report errors

Currently it just exits on the first sign of trouble but I think the best action is to continue running as many nodes as possible and then provide a status report at the end of the run.

Unless --throw is specified of course which should always bail immediately.