I recently came across a paper, "The Quantitative Trochee in Latin" (by R. Armin Mester, 1994) that seems to argue that feet in Latin were "strictly" bimoraic.
The arguments that Mester gives for parsing HLL words as (H)LL rather than as (HL)L seem relatively convincing to me (although it seems that Lahiri, Riad and Jacobs (1999) argue that Mester's analysis is incorrect).
But if trimoraic feet are completely unallowed in Latin, as Mester seems to say, then how can we account for the existence of LH words? The paper has a long discussion of "Brevis Brevians", which operated to replace LH with LL, but Mester acknowledges that it "was always an optional rule" (p. 12) and "was lost as a productive process in the later classical language" (pp. 12-13).
I'm confused about how Mester (or someone else who views Latin feet as maximally bimoraic) would account for the existence of LH words. Mester describes (LH) feet as "a gross violation of the quantity-prominence relations that are characterized in universal foot theory" (p. 14); but are they nevertheless supposed to have been tolerated in this particular context? If so, why exactly?
Section 4.1., on enclitics, proposes that stress may have been assigned to unfooted syllables in Latin in certain special contexts (words with enclitics), but that seems like a pretty "last resort" kind of analysis to me, and even Mester seems to treat it as somewhat dubious, saying that the "Mūsáque"-type stress pattern in Latin, even if it really existed, may have been artificially induced by contact with Greek (p. 54).