Excessive source masking in outlier files?

[schlafly]

Roughly 2% of brick-ccd combinations in bricks starting with 270* show a lot of masking of sources in the outlier rejection stage. Here is a particular example:
outlier-mask-2704p310.fits.fz decam-769920-N4

This example also shows that many of these have astrometric issues in the outlier-mask astrometry. I have checked that the real image WCS, however, don't seem to have this astrometric issue.

It would be good to understand and reduce this kind of "outlier" masking in DR10.