pster wrote:But I had never stopped to notice this case where there are three short vowels sloshing around and I was wondering whether there was any principle at work. My curiosity was also piqued by the fact that otherwise, this verb follows a general pattern.
τρέπω is a special instance because it has an /r/ in it. The three primary grades (as they're called) of ablaut are zero-grade, e-grade and o-grade. As you have observed, you often get zero-grade in aorists (and o-grade in perfects). Sometimes this is less obvious. For example, λείπω becomes λιπ-. That hardly looks like a zero root, but what has happened is the stem really ends in a consonant cluster, /jp/ (where /j/ is the IPA for an English "y"). In the zero-grade, the consonant /j/ turns into the vowel /i/.
What's going on in τρέπω is that the zero-grade ends up as /tr̥p-/, with the /r/ acting like a vowel in proto-indo-european. In fact, all the resonants, /r/, /l/, /m/ and /n/, could act as vowels in PIE. But, Greek didn't care for this, and so resolved these differently. /r̥/ and /l̥/ usually became /ra/ /la/ or /ar/ /al/ in various circumstances, with different dialects taking different paths (Attic καρδία, Homeric κραδίη in certain formulas, Cypriot κορζία). And some dialects went with with /ro/ /lo/ or /ol/ /or/.
In any case, ablaut did lead to the initial change, but /a/ is not an ablauting vowel, but the result of special circumstances for syllabic resonants.