"Quality" is "the asker's ability to receive an answer". We close questions that need focus, lack details, or lack MREs because we don't have enough information to answer them. When that happens, it is important that the question remains closed so that bad answers that lack context do not flood the question page.
"Structured downvote reasons" already exist. They're called close reasons. The only reason to switch from close reasons to "downvote reasons" would be to make downvotes less unfriendly, which I don't think will work. The perception of unfriendliness comes from the big negative number; adding text to that probably won't help. (Which we know because "on hold" didn't really work.)
Opinion-based
This is potentially fine in theory.
In practice, it will be bad for all the reasons that the experiment is bad. I am not going to rehash them here; there are way too many and (given company track records) I have insufficient faith that they would be addressed.
Duplicate
Duplicate questions still need to be closed. Otherwise, people will only notice the one transcluded answer from the canonical and inadvertently post duplicates of the canonical answers that weren't transcluded. If that happens, duplicates will stop serving their purpose: reducing duplication by pooling answers in one place.
Incorrect duplicates are not a "hard lock". There is already a recommended pathway for resolving them: editing the question and voting to reopen.
"[A]n author does not accept the proposed answer" is a very bad feedback signal. There are too many confounding variables: the author could've lost interest in the problem, solved it independently, lost access to their account, etc.
Off-topic
There are a lot of off-topic close reasons. How exactly do you plan on expanding them out? It seems like any time gained in clicks will be lost in scrolling.
Migrations by automated systems or non-moderator users seem like a really bad idea. We try to avoid migrations in general since experienced Stack Overflow users often do not understand the asking guidelines on other sites. In other words, I wouldn't trust the "smarter backend routing" to not. Migrate. Crap.
