> I also asked about the reduction of unstressed /a/ vs. /o/
> in Latin. Is /a/ more resistent to reduction than /o/?
No, but after the reduction analogy works differently.
All short vowels in medial open syllables act the same way. Normally
reduced to -i-, but phonetic conditions may produce something else, e.g. -e-
in the presence of /r/, -u- in the presence of velar-l or w.
However in compounds based on simplexes with -o-, the -o- was normally
restored, and this is not true of compounds with -a-. The reduction of -o-
to -i- survives only in forms whose derivation was not obvious, such as
memini < *me-mon-ai.
In closed syllables, the reduction of short vowels is to -e-, which
normally survives.
> So if *-osyo was reduced to *-ijjo/-i:o, *-asyo-
> should have done the same. Is that correct?
No. The first step is *-osyo, *-asyo > *-esyo. The -e- should then
survive, since the syllable is closed. Neither would be reduced to -ijjo.
Peter