I no longer think this analysis is correct; see the comments section below.
It’s well known that Quenya nouns ending in -a have genitives ending in -o, such as: ciryo genitive for cirya “ship”. This is because the primitive genitive suffix was -ō, and aō > ō in Quenya (which then shortened at the end of words as all long vowels did). Presumably this would also be true of nouns like fea “spirit” that ended in a cluster of vowels, whose genitive is probably *feo. But what about nouns ending in oa like loa “(seasonal) year” or coa “house”? A naive change of a to o would produce forms like löo, but Quenya does not tolerate identical vowels in hiatus.
There is a discussion of PE19/63 that describes developments of öo in two scenarios: (1) after the disappearance of intervocalic w in combinations like awo > öo > uo via dissimilation and (2) after the disappearance of intervocalic g in combinations like ogo > oʒo > o-o > ō. The first development must have been early (Ancient Quenya), because it predates intervocalic w > v, and the second development was therefore later. The question is, did aō > ō in genitives occur before or after öo > uo?
PE19/63 seems to provide an answer. In particular, it also discusses the change aō > ō, saying that it occurred after the loss of intervocalic ʒ < g and thus well after the dissimilation of öo > uo. The probable development of the genitive of lawa > loa is therefore: lawa-ō > loaō > lo-ō > lō. Thus the “naive” rule of a > o was probably correct in all cases, with the caveat that o-o > ó, and the genitives of loa and coa were probably ló and có.