I do understand that’s what you want but I’m trying to explain why we can’t.
In the scenario I mentioned, on Bear’s database, N1 has both #a and #a/b tags regardless of what tags appear inside the note (can be either #a/b or #a #a/b). When we apply the hide subtags filter we can’t scan the note text (too slow) and we should base our reasoning on what we have on the database. What we have doesn’t allow us to distinguish between N1 and a note containing #a #a/b in its text.
What you want can be done by changing how subtags work: N1 has just the tag #a/b on the database. But this will drastically change how subtags work for anybody, not just the people using hide subtags.
What you propose also breaks some other user expectations I mentioned before, but It’s pretty clear we don’t agree here and the technical reasons are more relevant.