In the analysis done on
DM-16875, the suggestion was made to override the default config settings for the source selection criteria for consideration of a given source as a suitable candidate for the PSF modeling. In particular, the default setting of a minimum flux of 12500 (raw counts, I believe) seems way too high and, ideally, the cut would not be based on flux, but rather on S/N. Justification for this based on HSC data as well as the S/N thresholding functionality was added on DM-17043. [Note that the fluxMin value has long been overridden to 4000 in obs_subaru].
It seemed this change was widely agreed upon by those involved in the imsim/DC2 Run2.2i processing and was adopted – in addition to several other source selection overrides for other calibration phases – in at least some of the DESC workflows (see, e.g. the overrides in their "dc2/run2.2" branch run: https://github.com/lsst/obs_lsst/blob/dc2/run2.2/config/imsim/processCcd.py#L39-L42
and note the other overrides in that file). Unfortunately, during an investigation on PREOPS-885, it was noted that these overrides have not found their way into the master/main config files in obs_lsst, so they are not currently in effect for any processing based off the master/main configs. This situation should be rectified.
Some care may be required in the actual S/N threshold set and whether the other updates in the file pointed to above should also be adopted. In particular, I would love to hear from anyone involved the DC2 processing campaigns as to the validation of those choices (I added a few watchers as a first guess, but please do add/remove as you see fit!) I also have no idea what was used in the DP0.1 processing run, so if anyone could chime in on that, it would be greatly appreciated!