Scaler will try to figure what to do with those duplicated UUIDs and from memory will try to reassign a different one on load to avoid clashes with existing content.
This operation reading and modifying files on disk would definitely slow down Scaler enough to be noticeable. We will investigate on our end if we can improve on this, but it is a bit of an edge case.
There is an old post (Scaler 1.2 ) with some info on the format that is still valid:
) with some info on the format that is still valid: