There's a term that, as far as I know, goes back to traditional Celtic grammar called "enumerators". These are essentially words that inflect for number in weird ways when preceded by a numeral, that are predominantly (but not solely) measure terms. For instance, Scottish reportedly disallows number marking for measure terms, meaning you can't say "three days", but rather "three day". Similarly, there are cases of nouns disallowing classifiers in classifier languages, which is presumably a similar phenomenon.
Two questions on this:
Do your pet languages display this feature? That is, is there a class of nouns that do not display normal number marking / classifier cooccurrence patterns? I'm aware of Japanese, Bengali, perhaps Thai, perhaps Vietnamese, perhaps Burmese, Dutch, Irish, and Scottish.
If so, are they only "measure terms", or do they extend to some weirder nouns?
A second, tangentially related question:
Are you familiar with any languages that have a syntactic mechanism for expressing a meaning like "approximately"? Specifically I'm looking for things that change the word order. I'm familiar that such a thing exists in Russian, where you can invert the order of the numeral and noun to mean approximately. So, "three cats" means what it means in English, but "cats three" means something like "approximately three cats". Furthermore, is there a limit to how high you can do this?
Ideally, I'd find a language that does both :)