Unless there is a logical contradiction in this phrasing, the proposed solution does not improves scalability:
 - "Bitcoins lasting forever" implies "unscalable";
 - "not prevent Bitcoin from lasting forever" implies "Bitcoins lasting forever";

 - Thus: "not prevent Bitcoin from lasting forever" implies "unscalable".

In practice, the only Bitcoin lost would be those whose owners forgot about or has lost the keys, because everyone with a significant amount of Bitcoins would always shift them around before it loses any luster (I wouldn't bother to move my Bitcoins every 10 years). I don't know how to estimate the percentage of UTXO is actually lost/forgotten, but I have the opinion it isn't worth the hassle.

As a side note, your estimate talks about block size, which is determines blockchain size, which can be "safely" pruned (if you are not considering new nodes might want to join the network, in case the full history is needed to be stored somewhere). But UTXO size, albeit related to the full blockchain size, is the part that currently can not be safely pruned, so I don't see the relevance of the analysis.

--