While investigating for
DM-24277, I learned that the stack does not apply the refcat proper motion data during the single frame AstrometryTask. The only task that currently passes the exposure epoch when loading the refcat is DirectMatchTask, which is only used for photometry. I think the solution is to pass the epoch in RefMatchTask, but I will have to check for other places that use loadSkyCircle and/or loadPixelBox to be sure.
It looks like we have no direct unittests of RefMatchTask, so any tests of this will have to live in tests of some higher level task (MatchPessimisticB? AstrometryTask?). We'll have to see whether we have an appropriate test refcat, or else create, extract, or mock one to have the relevant fields (at minimum, pm_ra, pm_dec, pm_raErr, pm_decErr).