An implementation of the proposal currently up on RFC-407 has been pushed on the branch for this ticket. It has not been fully tested. This ticket will stay "In Progress" until the RFC closes, and it's possible I'll add an attempt to address problem (3) on the RFC as well.

Changes are ready for review. I'd like to split it up as follows:

• Nate Pease: daf_persistence
• Hsin-Fang Chiang: pipe_base, pipe_drivers, pipe_tasks (recommend approaching them in that order)

Paul Price, while I don't want to burden you with a formal review while traveling, since you're the main user of the clobber/overwrite options, I want to make sure you have a chance to weigh in here as well. I'm most concerned that the new behavior (described on RFC-407) is acceptable to you; you're welcome but not expected to verify that the code here accomplishes that. If you do want to take a look but won't be able to do so immediately, just let me know so I can hold off on the merge until you have a chance.

Finally, Nate Lust should note that I've added a doDetection config option to multibandDriver that is now required to enable re-running detection. I believe this will only be needed when adding fake sources at the coadd stage, and it should be included in any instructions for doing so.

I will be writing a community post announcing these changes and describing their usage once we're ready to merge. I think it's premature to do that now.

daf_persistence looks fine to me.

Sounds good to me. I'm glad you've found a way to reduce these long-annoying problems.

Looks good to me. Two comments on the PRs.

Merged to master.