See implementation of `getLatestNonceGivenHighestFinalNonce()`. Most probably, we can adjust / optimize it. Currently, it's a bit conservative.