Actually the forms you quote (except for *oid-/*id-) fit much better into
the known patterns if one DOESN'T reconstruct them with *h3-.
First, Pokorny's reconstructions are anti-laryngealist. One
consequence of that is that Pokorny reconstructs things like "diphthongs" *@i
and *@u, which make no sense in the laryngeal theory (a prevocalic laryngeal
cannot be vocalised). In laryngealist terms, forms like *aus- should be
reconstructed with a full vowel, and the colour "a" in Latin and/or Greek points
to *h2-.
The "ear" root is thus reconstructed as *h2aus- (= e-grade
/h2eus-/), with the corresponding o-grade presumably surviving in Greek (it
may reflect the original root formation *h2ous- as opposed to derivatives
like the popular es-neuter *h2aus-es- or Greek *h2aus-nt-).
Again, the "bird" word gives Latin avis, which points to
*h2awi- (an archaic Gen. form is visible in Sanskrit ves. = h2weis). Greek
aietos < *awi-eto- provides more evidence for *h2- rather than *h3-. The
"egg" word (usually reconstructed as *o:wiom/*o:(u)jom) may be related ,
but the details are not clear; some etymologists opt for a vrddhied derivative,
others for a compound. In any event a form that is not fully understood cannot
be used as evidence.
The "mouth" word poses special problems. My private (but
published) opinion is that the word is an old athematic neuter *(h1)o:s,
*(h1)es-ós, reanalysed as an es-stem: *o:s > *ó-es or *ó-os, hence the
innovated oblique forms *(o)-és-os and (younger) ó-es-os > *o:sos. This
explains both the non-Anatolian forms and the Hittite development (as analysed
by Szemerényi 1978: ais, issas < uncontracted *o-es, *(o-)es-os). The absence
of a laryngeal reflex in Hittite points to *h1- (which may also mean that there
was no initial consonant at all in PIE). The Slavic form (if related!) may
have been affected by the "ear" word.
[I'll answer tomorrow regarding Latin o:di: and "Attic" redupliaction --
it's getting late.]
Piotr