I know nothing about Armenian, Old or New, apart from vosp, ’cause I like lentil soup.
I stared for half an hour at:
- A Grammatical Sketch of Classical Armenian,
- Damme, Dirk Van – a Short Classical Armenian Grammar
- Grammaire armenienne : M. Lauer and A. Carriere .
I think I have the answer.
Old Armenian does not have nouns whose nominatives end in a vowel. So the a-declension, those nouns that in Latin and Greek ended in -a, end in a consonant. Greek gynē corresponds to OArm kin. It seems that final unstressed vowels were systematically chopped off in Proto-Armenian.
So, if a Greek word like plateia comes into Armenian (via Syriac plāṭīā), “square, public street”, Old Armenian could not deal with it as a nominative: it wouldn’t fit the patterns. (I don’t know how are supposed to work when your plural ending is -kʿ: I mean, sg.nom. azg, pl.nom azgkʿ ? Seriously?)
In Old Armenian, plāṭīā ends up as połotay.
Account #1. To make it fit, the word has to end in a consonant. Chopping off the vowels wouldn’t work well, you’d end up with plat, which doesn’t sound close enough to plāṭīā. So the safe thing to do is to add a consonant to the foreign word. And –y is the best consonant to add, because it’s a glide: the result still has a similar syllable structure to the original.
I see that Greek hylē is borrowed as հիւղէ (hiwłē). But Wiktionary also notes the variants hiłeay hiwł and hiwłay, so there was a strong trend to go with -ay.
Account #2. But it may be that this is just a trick of orthography. Lauer & Carriere say that final –ay is pronounced –ā. So this could just be that plāṭīā was pronounced połotā in Armenian, and –ay was how Armenian wrote down the new-fangled long final a.
In any case, this looks like stuff internal to Armenian.
Again, this is all extrapolated guesswork from a couple of sketch grammars.