I'm not fully sure how matching releases works, but I know that regularly Picard doesn't automatically pick the correct release. F.e. because it prefers to match on track count even if the tracks are duplicates but also it strongly prefers to match by track length even if the tracks only differ by 5s and the track title are an almost perfect match.

Instead of trying to fix/improve all these types of matching logic, it would be nice if there was a reverse option where Picard just loads all known releases for an album and selects the release that has the least number of duplicate and/or unknown files.

While working on this project: https://community.metabrainz.org/t/feature-suggestions-after-matching-a-huge-messed-up-folder/712286 this is basically what I was doing like half the time, just trying different releases to see which one matched the best. And I think that in at least half of those cases just selecting a different release did really work.