Bingo. I think this is very important insight from CS to this kind of literature. You might enjoy Robert Bringhurst’s discussion of Haida myths, where he calls them fractal in a way that doesn’t set my teeth on edge (the way most non-rigorous uses of that word do).
This is the kind of thing that makes me sad about the two cultures problem. The few scholarly discussions of oral literature that I’ve read seem to be groping for this kind of concept, while a lot of hackers I know are dismissive of the idea that there’s anything of real interest in fields like anthropology. Everyone loses if we can’t find ways to say things like “self-similar structures compress well” across fields.
Edit to self-link: I tried to make this kind of idea engaging to some of my friends at http://basecase.org/env/time-and-myth . The book where Bringhurst talks about this: http://books.google.com/books?id=QhMkb1HOjz4C .
It’s not just that we don’t think of “rosy-fingered dawn” as being fresh because we know Homer used it: it’s that even if you’ve never heard it before, it will have lost its freshness halfway through the Odyssey.
So word-by-word freshness is usually not a useful or fun thing to look for in epic literature. While people can criticize Tolkien for a lack of eye-kicks (as I think I remember William Gibson calling phrase-scale fireworks in cyberpunk), a tree can be green-fingered twenty times without it being a violation of the rules of the game he’s trying to play.
(Also, for all we know, Homer might have stolen all his clichés.)