To me this is sort of like the ‘usage is the rule’ concept in language. I’m all for the addition of new words to the language, but not so enthusiastic about the deterioration of rules and structures, especially when it makes communication somehow clumsier or less efficient. Though this doesn’t mean that both processes are unavoidable, it means that I’ll do what I can to slow down the latter.
I think we agree on this more or less; the rules of music notation are for the most part intended to preserve clarity and effectiveness. We would like to slow down the deterioration of rules and systems, as well as facilitate the addition of new methods and notations as they are created.
However, I don’t know of any rule or guideline suggesting that tuplets in grace notes are NOT allowed, nor do I really see why they shouldn’t be, if our goal is clarity and efficiency.
Performers certainly play notes in triplets differently than notes without — Gérard Grisey has been known to make extensive use of this, but I think it probably goes much farther back (I wish we could search IMSLP by note value as well as motif). So why shouldn’t it extend to grace notes?
In the piece in question (Café 1930 in Histoire du Tango), Piazzolla makes a distinction between three grace note 16ths and a grace note 16th triplet, as well as between quintuplets of the same. So unless the conclusion is that Piazzolla and his editors were haphazard or unobservant enough to miss them, these were intentional differences.