Would you mind giving this a look?  Seeing as you have been poking around the plots from pipe_analysis, you understand the motivation.

Unit testing is not really feasible as we need a fully processed (through multiband) dataset, so I added a test in ci_hsc.

The PR for pipe_tasks (which is not showing up in Jira these days) is: https://github.com/lsst/pipe_tasks/pull/217

I ran a Jenkins run before the weekend which passed. Just to be on the safe side, I also kicked one off today [UPDATE: it also passed].

Oh, and I also created a coadd and checked that the sources that get these flags set do indeed look sensible.

Looks good to me. The only thing I wasn't clear about is why the default values are all 0.2. If you could possibly clarify that in the docstring, that would be helpful.

Very fair question.  There is a lengthy discussion in the task description here: https://github.com/lsst/pipe_tasks/blob/master/python/lsst/pipe/tasks/propagateVisitFlags.py#L64-L90

This seemed like more than enough justification to me, and I felt the logic also applies to the flags added here.  Let me know if you disagree or think a note needs to be added elsewhere to make it more visible.

Indeed, that provides plenty of justification, but is a rather odd place for it. I think I was further confused by the default for the following config field, matchRadius, which also happens to be 0.2... arcseconds.

I'd prefer a second sentence added to the docstring along the lines of, "May be in the range 0-1; default 0.2 will flag the coadd if at least 20% of its inputs are flagged." I didn't fully understand that's what was happening here until you pointed me to the task description.

The config parameter doc now reads (with a --show config=propagateFlags* on a multiband.py run):

# Source catalog flags to propagate, with the threshold of relative occurrence (valid range: [0-1], default is 0.2). Coadd object will have flag set if the fraction of input visits in which it is flagged is greater than the threshold.

config.measureCoaddSources.propagateFlags.flags={'calib_psf_candidate': 0.2, 'calib_psf_used': 0.2, 'calib_psf_reserved': 0.2, 'calib_astrometry_used': 0.2, 'calib_photometry_used': 0.2, 'calib_photometry_reserved': 0.2}

Thanks Meredith.  Docstring updated and merged to master.