This was discussed at the AP Pipeline Meeting of 2020-04-06. At that meeting, we theorized that this commit on
DM-23616 might be responsible.
In fact, the change does come from
DM-23616, but not that commit. Rather, DM-23616 had the side effect of changing config.ccdProcessor.calibrate.astrometry.wcsFitter.order from 2 to 4.
In more detail: obs_decam contains files config/processCcd.py and config/processCcdCpIsr.py, where the latter is now obsolete but used to be specific configuration for using cp calibration products. ap_verify_ci_hits2015 used to refer to the latter, but now refers to the former. processCcd.py sets calibrate.wcsFitter.order=4 (actually, it used to do this directly, but now delegates to calibrate.py which does it instead). processCcdCpIsr.py did not (although now it does, since it now delegates to processCcd.py, which, in turn, delegates to calibrate.py).
Next question obviously is: what should this be set to?
RFC-577 reduced it to 2 by default. Despite a comment from John Parejko on that ticket that “a 2nd order fit is probably what we want”, DM-18293 implemented RFC-577 by overriding that default to 4 for obs_decam (except not in the case of CP calibs, for the reasons described above). So DM-23616 effectively enabled an earlier override that had been inadvertently disabled.