Restarting a job overwrites the existing output file

Care has to be taken when restarting as the new output will be written to a file matching the existing output and hence we clobber the original data. Ideally we would simply append to the existing data (but there are common cases where we do want to clobber, i.e. when rerunning a case but not restarting).

This is probably something to be considered as a part of a longer term/large scale look at addressing all of our I/O.