@1 This is a good solution I think because it is already part of Calibre. At the other hand, I already have a lot of this searches, but that is a personal thing, solution works for me.
I've a large group indeed.
I exempt the books that gave a problem previous (did not yet rename them)
I exempt books I previous marked as not duplicate (put [other version] in title)
So at the moment there are 269 books exempt (no need if I use solution for 1)
The script is (even fast (I have 2 pc's, even on my old pc it is a fast process, with more exempts it is slower) So I think a complete test would be no big problem.
To solve the problem maybe you could use the following workflow (do not know how it is implemented at this moment):