I agree that it can be compressed. The interest in exploring the applicability of commitments is that this scheme still leaves UTXO cost O(N) just with a better constant factor.
It’s interesting to consider how that could evolve over time e.g. once the reduced cost justify doing something about it. Like the utxotree stuff results in a kilobyte of state, which 800MB (plus the non-dust utxo set) is very large against indeed. Particularly where someone may want a circuit that validates (most of) a block that difference matters. But the more efficient solutions have the problem of updatability, while the txo suggestion needs only a static proof which is a big win.
At least I’ve always found it useful to try thinking a couple steps ahead of any proposal, sometimes it results in changing how I think about it.
That seems somewhat dubious to me, I could argue for that side but it feels like a stretch.