Please i just want to verify that i well understood : you are saying that NOT EXISTS is better than MINUS / MINUS ALL , that's it ?
Can you give me more details please ?
As Dieter said (years ago): MINUS / MINUS ALL will always spool / redistribute / sort both tables.
If you code a NOT EXISTS, the optimizer will take advantage of the PI being the same and just do a direct AMP-local join.
Note that the results for NOT EXISTS are equivalent to MINUS, not MINUS ALL. That's fine for SET tables, but not for a MULTISET table with duplicate rows.