Many users of CJK language would argue that CJK unification was a mistake.
Luckily it's not a decision without turning back. In most relevant contexts you should know the input language and can select a Font specifically using said variations. Of course this information will not be present in plain text, but if it turns out to become an issue I'd wager, since language codes do exist, that a control code-point for language selection can be added to the specification. There's already so many special cases in Unicode that it shouldn't be a huge issue (apart from backwards-incompatibility that would lead to tofu instead of no rendered glyph).
Comments
Luckily it's not a decision without turning back. In most relevant contexts you should know the input language and can select a Font specifically using said variations. Of course this information will not be present in plain text, but if it turns out to become an issue I'd wager, since language codes do exist, that a control code-point for language selection can be added to the specification. There's already so many special cases in Unicode that it shouldn't be a huge issue (apart from backwards-incompatibility that would lead to tofu instead of no rendered glyph).