Older HSC files can have incorrect DATA-TYP headers, causing some science frames to be ingested as calibrations and probably vice versa.  The OBJECT headers seem to provide values that could be used instead if we can completely enumerate the values used for calibrations; given the finite time duration of the problem.  Hisanori Furusawa also reports that the STARS database at NAOJ should be updated with correct values for these, so we could also query that to obtain more complete corrections for all potentially-incorrect files.

Priority for this ticket should be to get in a fix that enables the correct Gen3 Butler ingestion of all affected science frames in the RC2 dataset, which I can provide a complete list of later.  We can create a new ticket for a more thorough fix if the solution here is more localized.

Tim Jenness, please steal this if you think you'll have time to work on it today.  If not I'll be bugging you about the best way to approach putting the corrections in astro_metadata_translator.