Considering written languages that we know, restricting ourselves to alphabetical languages if helpful: Can we make any general statements or assert any constraints on how a language could have ever been spoken? (I think this would be of particular interest when considering a dead language like Latin.)
For example, can we place constraints on the number of syllables or phonemes that might have ever been used in established dialects for a particular word? Consider the French "hors d'oeuvres," in which (strangely, to this English speaker) the second word is commonly pronounced with a single syllable. I can imagine it being pronounced with up to four syllables, but no more than that. Or consider the English "one:" I've heard rural dialects in which it is spoken with two syllables ("wuh-uhn"), but I don't imagine it could ever have been spoken with more than two. In terms of phonemes: I would be surprised if the English word "one" were ever spoken with fricative phonemes. These constraints that I am imagining are intentionally very broad; I am wondering whether linguists have found any useful constraints, and if so whether they get significantly more narrow.
Or are there other constraints that can be articulated for spoken languages? For example, perhaps there are phonemes that we know have never been used in particular branches of particular languages?