Since the acceptance of
RFC-356 and RFC-289 (and maybe another I haven't found?) selected maggies as DM's calibrated flux unit, the Project Science Team has landed firmly on janskies (Jy) as the unit for calibrated fluxes in public LSST data products, with a further note that nano-janskies (nJy) put many relevant quantities for LSST in a nice range for humans to reason about and remember. For more information, see https://pstn-001.lsst.io/.
I can't think of a good reason for DM's internal calibrated flux unit to be something different, and I can certainly imagine ways in which being different could cause problems (probably minor ones, but then there's the Mars Climate Orbiter...).
So I hereby propose that all DM code and intermediate data products use nJy for calibrated fluxes, and that we drop usage of maggies entirely.
This does not mean that we should convert instrumental fluxes in images or catalogs to calibrated fluxes as soon as a calibration is available; on the contrary, we should work in instrumental fluxes and maintain the calibration (the mapping from counts to nJy) in a separate, more lightweight form (i.e. Calib and PhotoCalib) as long as possible. This avoids frequent re-calibration of images and catalogs as the mapping from instrumental to calibrated fluxes improves over the course of pipeline execution. We currently continue to use instrumental fluxes even through coaddition and coadd processing, and while I think I'm open to switching to calibrated fluxes earlier, that should be discussed on a different RFC (possibly RFC-545).