BIP-327 MuSig2 1.0.5 adds test vectors for implementation mistakes the old ones missed
The BIPs repository merged pull request #2351, "bip-0327: add more test vectors", on 16 November. It updates BIP-327, the MuSig2 multi-signature specification, to version 1.0.5, dated 10 November, with the changelog entry: "Add test vectors that detect implementation mistakes the previous test vectors missed, and improve a PartialSigVerify test vector." The pull request was opened by Jonas Nick.
Each commit names the mistakes its vectors catch. Sign vectors with the second secret nonce value k2' equal to n and to n + 1 detect implementations that check secret nonces only for zero, that check k2' <= n instead of k2' < n, or that reduce k2' mod n before checking it. A vector with a secret nonce generated for the negation of the signer's public key, and a KeyAgg vector with a key and its negation, catch code that compares public keys only by their X coordinate - the two share it and differ only in the first byte of the compressed encoding. A NonceGen vector whose random input and secret key are not a repeated byte, and whose public key has an odd Y coordinate, detects reversed byte order and code that uses only the first byte of the random input; in the earlier vectors those inputs were 0F and 02 repeated 32 times.
One PartialSigVerify vector is changed because, as the commit explains, the partial signature n failed the verification equation even without the range check, so it could not detect code that omitted the check or tested s > n instead of s >= n. Two tweak vectors catch implementations that set the accumulated sign gacc to g instead of multiplying by it. Further vectors cover a single signer, a 4-byte extra input and both tweak types in DeterministicSign.
A new README sets the rule: the vectors contain only inputs that meet the lengths and counts the BIP specifies, and a vector "should only be added if it detects a plausible implementation mistake that the other test vectors do not detect", with a comment naming that mistake.

Why it matters
A test vector only proves something if a plausible bug would fail it. This change starts from that question for each vector, and the commit messages show several cases where an existing vector passed for the wrong reason.