doesn't make much sense to use quantum-resistant hashing along with ECDSA
Why not?
1. As long as ECDSA is safe, it will work, and will take small amount of on-chain bytes.
2. When it will be unsafe, then spending only by ECDSA can be blocked, in exactly the same way, as spending only by key in P2TR can be.
3. If some quantum proposal will be broken classically, it will be possible to downgrade it, and old nodes will still understand it.
4. It is strictly connected with sigops limit, so you can handle the same number of signature checking operations, as old nodes did.
we can completely ignore old nodes and simply "kick them off" from the network
Sure, but it shouldn't be your starting point, but rather your end goal. First, things should be optional, because if they will be mandatory from the start, then it will be quickly turned into hard-fork, and will end in the same way as BCH, and other forks. So, the first thing to do, is to get a compatible client, which will work on the same chain, as existing clients. And because they cannot understand bigger blocks than 4 MB, then new signatures should be somehow attached to the existing ones. And it is much easier to attach them through R-value, than through additional OP_RETURN, because then, it takes less on-chain bytes, and can provide the same security (if someone can fake R-value, then that person can fake OP_RETURN entry as well).