Actually, I would expect Windows to find the UTF-8 escape sequence in the supplementary volume descriptor, refuse to mount data in an unknown character code, and switch over to mounting the primary volume descriptor instead.
But that's by the way -- I don't particularly care for how systems fail at handling registered character set support in ISO-9660, or avoid doing it -- I was trying to ask for systems that succeed at doing it over and above the level of Joliet.
Linux mkisofs describes the situation fairly well under "CHARACTER SETS":
Looks like it's RRIP support only, and thus unspecified by ISO-9660.
Perhaps consider moving to UDF instead?
I seem to have formulated myself badly. My focus right now is on conformance testing of ISO-9660 CDs, and switching to another file system thus not an option.