Bids-app Running Feature to save work without all of the participant jobs being completed

[shots47s]

As we have found that many of the useful "BIDS" datasets are not standard, it is difficult when launching a BIDS-app to distinguish that when a task fails, it is because of machine error, application error, or that there is no data. Currently, if one setup up the job, it will only save if all of the participant applications have gone to the "Completed" state. We discussed a few options on how to resolve this, and I believe this is what we decided.

We will implement the Save task to be able to save all of the jobs that have made it to the "Completed" state.

Do we:

Make this an optional feature?
How does restarting the failed jobs work in this scheme, can we easily add them to the dataset when they are restarted?
It would seem then that we allow group computations over partial datasets (think we have to).
Putting here to discuss so that Pierre has a clear roadmap.