> Yeah, I think that would work. Obviously best to test in a throwaway> database ...
Thanks, Tom. Fixed it with that query.
As for the proper fix, indeed it looks like the 'drop role' needs to be taught to update pg_init_privs.
-Floris