I just did a quick read. The idea is fun, but probably not compatible with mimvlewimble chains since you need all UTX0’s to prove non-inflation.
I am quite certain that if it ever will be needed some compaction of kernels for spend UTXO’s can and will be done, e.g. inclusion proofs. So Grin already deals nicely with spend UTXO’s and chain growth and can probably even further reduce the size of spend UTXO’s if needed. But I think Grin will always need to store the full UTXO set with range proofs. For now there is little need since Grin is already so minimal🙂.
A I thought I red about it before, utreexo was discussed in 2021: