This is a classic problem and I'm not sure I expect a good answer to it, but it's worth it anyway. The question is partly about what appears to be some specious reasoning around Verner's Law forms and secondarily about the a-stem plural in northern West Germanic languages (OE and OFr).
We'll start with the a-stem genitive singular, which has an -s, unlike the genitive singular of all the other noun classes. Ringe says essentially that this is due to influence from the pronouns, such as *hwes, where it would be regular. Gothic implies as much since the a-stem ending is exactly the same as the pronominal ending and different from what we'd expect (-is instead of ˣ-as). This feels satisfactory to me.
Sihler, on the other hand, says that Germanic essentially requires that *-osyo be reconstructed for o-stem genitive singular. His reasoning is suspect. He denies, without good argument, the traditional explanation. He then rejects *-oso because it would have been voiced due to Verner's Law, but is voiceless in WGmc. The only solution to this problem, according to him, is to reconstruct *-osjo, from which PGmc *-asja, wherein the *j somehow prevents Verner's Law. This is not explained at all and, aside from some short commentary on OCS, that's the end of the section.
This same kind of argument about Verner's Law is also invoked to explain the NW Germanic a-stem nom pl. Instead of coming from PIE *-ōs, which would produce *-ôz in PGMc and ˣ-a in OE, it must instead come from something like *-ōses. I would expect that to become *-ôziz and then something like ˣ-ur in OE, neither of which are attested at all. It is implied that it would have been *-ôsiz, which would regularly produce the attested -as. What is the explanation for the voiceless medial /s/ instead of expected /z/? If it's because we might see some nouns with stress on the ending, triggering the voiceless Verner's Law variant, then we are back to square one. We could just as well have the simpler ending with the voiceless variant, which is actually attested in other Gmc languages.
What am I missing here?
