I've done the last feature, bringing it to its fullest extent in
Late Andanese. There's not a whole lot of grammar info on that page, but if it can somehow help even a little bit, Im glad to help out. I should note, though, that Late Andanese is not intended to be naturalistic, and I think not even Swahili has quite so many noun classes as Andanese. More importantly perhaps, I dont think there is any natlang in which the same root can appear in a dozen or more noun classes. but i like it, and i especially like the wild semantic drifts I've come up with such as
himaku "money" ~
mimaku "poison" just by changing the noun class prefix.
Some Bantu languages, such as Bemba, mark noun case with a classifier prefix as well, which I think means that the normal classifier is omitted, and words merge. I dont do this in Andanese because there are so many roots that are spread across multiple noun classes ... e.g. if I did what Bemba does, i would have no way to distinguish between "to the book" and "to the school" since both words use the same root. You could do something like this on purpose if you want a quirky grammar, but my guess is that the speakers would just stop borrowing words across multiple noun classes. My solution is to just stack the case markers on top of the classifiers, which gets a little clumsy but I'm okay with it.
i think Pirahã might also get 3rd person pronouns from the syllables of the antecedent, which is similar, but there's not much info about Piraha online. I dont think Piraha has proper noun classifiers.