I'm not sure the above calculation is ideal. Let's talk about what we want here. With the fees dropped, this makes a 1 input + 2 output transaction cost 1sat/byte roughly. However, I don't like the linear formula, and the 100% return rate when spending a UTXO.
Since spending a UTXO doesn't increase the UTXO set size, we should be safe here. I think the linear formula will work for now and can be optimized as we continue to improve the mempool and other more critical paths.