Nice, that was quick. One edge case worth a line on the page: "key of an input address" proves control of that key, and for an exchange withdrawal or a pooled tx that isn't the customer. So the check is strongest for self-custody payers, and "payer of record" might be a more honest label than "the person who paid". Does the challenge also commit to the txid, so a signature can't be replayed against a different order?