Seems like there's a real need that this is aimed to address. I don't understand how this protocol prevents double-spends though. If a payer sends the
same signed payment authorization (i.e. same nonce) to two different online services at the same time, I think they would both see the payment as valid. Then both services will submit the same nonce to the blockchain but only one can succeed, the attacker gets double the value while paying only once when the first settlement transaction confirms and the second one reverts.
Maybe this could be mitigated by only accepting substantial payments from payers with a good reputation, since my understanding is that every payer's transaction history is fully public (which is its own can of worms).