I mostly agree with all your points, although I think the risk of a centralized takover will not be high in the near term, once the delayed stake voting is deployed, which removes the highest risk of a centralization attack via rapid custodial stake voting.
On a related note, my original question wasn't meant to identify any "increased decentralization". I just thought the original data showed too much whale stake as "inactive" which seemed suspect to me, as it does provide a pretty reasonable rate of return versus other investments right now, so I thought it unlikely that it wasn't being used to curate. And my theory, which is now well supported by the data, is that whale stake is in fact largely active, it's just via delegations rather than direct voting.
It appears you were quite correct.