So far it would appear that Classical Japanese had three C-stem conjugations (which I will henceforth refer to as CI, CII & CIII, corresponding to the traditional yodan, kami-nidan and shimo-nidan, resp.) and one V-stem conjugation (traditional kami- and shimo-ichidan). The four traditional irregular verbs (ka/sa/ra/na-hen) can be analyzed as mixed-conjugation or minimally irregular C-stems. (E.g., ari would be a typical yodan verb if not for its non-characteristic SS (shushikei).)
Naturally there are a few wrinkles I’m still trying to iron out.
One of the most interesting things to turn up in the results is that the morpheme used to create the most basic (free-standing) form of the verb (SS) remains unchanged in the modern language, showing the same allomorphy between -u and -ru.
The details of other morphemes are a bit more problematic. Take amida’s favorite raru, for example. A selection of relevant examples would be
kiraru (keraru) (V)
Assuming allomorphy in the endings (as in the analysis of modern language presented earlier) results in four allomorphs: -ar
u (CI, na, ra), -rar u (V, ka, sa), -irar u (CII) and -erar u (CIII); this isn’t impossible, but I’m wondering about the justification. Allomorphy in the stem (basically the traditional analysis) seems like a lot more “overhead”, with at least three or four versions of each of the consonant verb stems (e.g., sak-, saka-, saki-, sake- for CI, sug-, sugi-, sugu- for CII) and still two allomorphs of the ending (-ru ~ -raru).
Another alternative would be to assume the existence of “linking morphemes” which must be added to C-stems before given endings (sak+a+ru); as with the above stem-allomorphy analysis, there still must be ending allomorphs.
Will continue to ponder and keep y’all posted….
(Addendum: Realized that the final -u on all the forms is the self same SS ending.)