I implemented BIP 360 in Bitcoin Core on regtest as an independent exercise and ran it against the official test vectors: all construction vectors pass, and the pqc vector file turned out to carry two bugs (fixes submitted as bitcoin/bips#2220, with doc fixes in #2221 and a correction to the claimed witness size saving in #2223).
Full write-up with measured witness sizes and some spec feedback: BIP 360 (P2MR) implemented in Bitcoin Core on regtest: vector results, measurements, spec feedback
One gap worth flagging here: the official vectors are construction-only, so consensus spend-path testing has no shared oracle yet. I’m happy to generate and contribute spend-path vectors upstream if that would be useful.