One observation about these out-of-band non-atomic accelerators is that the fee is earned/spendable immediately by miners. This is because the fee is not included in the coinbase output and hence not locked for 100 blocks the way that mining rewards are supposed to be.
It may not be something to worry about right now, but in the far future, if somehow most transactions end up paying their fees out of band, then the network security may suffer. Would be curious to get @petertodd 's take on this.
Interesting point! I hadn't considered the 100 blocks thing.
And yeah, in general it's quite harmful if out of band payments become a significant fraction of fee revenue. Realistically, only more centralized miners can profit from them. Decentralized mining like P2Pool certainly can't.
Thanks for your reply. Having fees be implicit in a transaction as opposed to existing as a separate output saves space in the block (and in the utxo set), but maybe in the future people will just pay their fees to an output of <k blocks> OP_CSV thereby emulating (with k = 100) the traditional incentives.
It also is interesting to think about what might happen if people started extending k to longer durations.
Rationally, for larger k, this "explicit fee output" would cost the sender more sats (time value of money) to get the transaction confirmed (assuming miners would even be on the lookout for such transactions with these outputs), but if the mempools were to have a lot of otherwise equal value transactions with different ks, then we might be able to get a near-real-time gauge of miner time preference/myopia.
<k blocks> OP_CSVthereby emulating (withk = 100) the traditional incentives.kto longer durations.k, this "explicit fee output" would cost the sender more sats (time value of money) to get the transaction confirmed (assuming miners would even be on the lookout for such transactions with these outputs), but if the mempools were to have a lot of otherwise equal value transactions with differentks, then we might be able to get a near-real-time gauge of miner time preference/myopia.