There are a few different versions of the FLUXSTD catalogs that we have used in our engineering runs. There is room for improvements in the latest version and this epic is a master ticket for the next (and hopefully final) version of the FLUXSTD catalog. The catalogs so far are summarized in
https://github.com/Subaru-PFS/drp_fstar_photo/blob/main/doc/catalog.md

Based on a discussion with Ishigaki-san, the current version is based on the Gaia DR3 and PS1 DR2 matched catalog. The catalog is stored in a local DB in Mitaka. The SQL script at the link above gives the selection function, but in short: gMeanPSFMag<21 and gMeanPSFMag>13.5 are the primary constraints. The comparisons with SEGUE included the y-band photometry, but Ishigaki-san told me that faint objects are missing the y-band photometry (need to check why) and most stars are classified without the y-band. We need to understand this because the y-band is useful (z/y-bands straddle the Paschen break).

Also, no flag cuts have been applied. We should apply, at minimum, qualityFlag & objInfoFlag == GOOD. There may be other flags to apply and we need to look into it. We should also remove bad stars; we may remove red stars (after the Galactic extinction correction is applied). The original catalog has about 2 million objects and the compute time is about 3 seconds per object.

Another thing we need to understand is how brutus behaves when we exceed a specified memory limit. Ishigaki-san says it uses fewer models without any log output.