I think this BIP does it wrong, even if the idea isn't bad and I'm all for ideas that mitigate the UTXO set spam a bit.

There are some of these UTXOs generated by these data/spam techniques that are unspendable, and these are those we should worry about. Not to make additional UTXOs unspendable.

But I think no consensus changes are necessary for this. Instead, nodes should be simply enabled to apply their own calculations, by themselves, which allows them to drop UTXOs they recognize as unspendable from their UTXO set. What method they use would be up to them, they could of course use the external tools or any other heuristics they think viable.

If there's a false positive, and a node drops an UTXO which is spendable, this would not be the end of the world. When that UTXO is spent and they receive the transaction, they would detect it first as invalid and not rely it. However, once they receive a block with the transaction, then the node would become aware of their "fault" and they could reconstruct the UTXO with blockchain data.

Basically it would be always a tradeoff: if their UTXO classification algorithm is too aggressive, then they would have higher running costs later to reconstruct the UTXOs and because they wouldn't be able to use the compact block mechanism for blocks containing that txes.