It’s my understanding that one-to-first joins on enterprise geodatabase data are non-deterministic. Despite what the Add Join documentation says, the related record with the lowest OBJECTID is not used. Instead, an arbitrary related record is used; the related record can change without an apparent reason. That's problematic from a data analysis perspective.
Related: Choose if join will be one-to-first or one-to-many
This idea is to make one-to-first joins be deterministic on enterprise geodatabase data.
