Percent-encoding is a mismatch and tracked by #796.

However, "valid domain" also has a bunch of rules that the host parser does not attempt to log validation errors for. We should either make it log those validation errors or remove this mismatch. (To be specific, "valid domain" uses a different value for "domain to ASCII"'s toStrict parameter and also invokes "domain to Unicode".)

(Then there's #245 about not defining this in an imperative manner, which would have made this harder to spot presumably. Hmm.)