Practical considerations:

1/ only taproot utxos

2/ address reuse is not supported (though it could be with a trivial algo change

3/ notice the anon set is not "all utxos in the keyset provided" (say it was "all taproot utxos above 0.1 btc" for example). if you specify 5 utxos and you prove "total is between 10 and 12 btc", then obviously not all combinations of 4 utxos add up to something in that range. So if you started with e.g. 300k, your actual anon set might be more like 5000 (guess!). (but even more confusing - 5k utxos might be *part* of a set of 4 that adds up to that range, but the number of actual combinations of 4 could be astronomically larger). Due to nonlinear distribution of utxos, it will therefore be better, if you have a large treasury, to split your coins into a lot more small values, and then build proofs of sums with larger utxo numbers.

4/ proofs in the current code are not aggregated, which makes proving for say 50 utxos, not 4, impractically slow. But this is an implementation detail: aggregation of these bulletproofs is a well established algorithm, and removes this problem. The verification here is very fast, as with aut-ct.

By the way if you're wondering where any documentation of this algorithm is, yet, don't, I haven't written it yet, it's on the way.