Well first of all a string does not equal a label, since there are further restrictions on the latter - I think a minimum length of 3 characters in the case of A-labels is one. Secondly, in the case of Arabic
LGR, we defined 16 different sequences which cannot co-occur in labels, only that we did that in the form of
WLEs rather than variants. But I seem to remember a discussion among the GP and in between the GP and
IP, where
IP explained that such
confusabilities can be dealt with either in the form of variant rules or whole label evaluation rules. Anyhow RFC 8288 gives a case example for variants of sequences exactly parallel to some of those cross-script variant candidates I was suggesting:
17. Variants for Sequences
Variant mappings can be defined between sequences or between a code
point and a sequence. For example, one might define a "blocked"
variant between the sequence "rn" and the code point "m" because they
are practically indistinguishable in common UI fonts.
Since we are discussing cross-script variants, I don't think WLEs will be able to control them, which would mean that we must deal with such confusable characters or sequences of characters in the context of variants, and therefore come up with a stronger criterion than"not a "single" code point in the repertoire"