Thematic *-e/o- : *-i- revisited

I'd like to share some thoughts concerning the behaviour of the thematic
vowel in derivatives. As already observed by Jens Rasmussen, a
morpheme-final *-e/o- is never deleted by word-formation rules but may
under some specific conditions be weakened into a high vowel. In
particular, the thematic vowel turns into *i if a derivational process
adds another *-e/o- to it, producing *-i-o- --> *-(i)jo-. Now, *i is
also frequent in suffixes like *-iko-, *-ino-, *-ito-, *-isk^o-, often
added to thematic base forms whose thematic vowel is allegedly
"replaced" by the suffixal *i.

Let's hypothesise that at a certain point in the development of PIE the
rule was (approximately) as follows: in derivatives, the final vowel of
a thematic base was reduced to *i if unaccented and realised as *e if
accented. Let's symbolise the thematic vowel as *E; we could then expect
the following
developments:

*-E-É- > *-ijo- (as in adjectives derived from thematic nouns)
*-E-tÉ > *-ito-

etc.

Substantivisation and other early accent-shifting processes could
produce forms with *e rather than *i, as in mass nouns in *-éjo-m (coll.
*-ijah2), or de-adjectival abstract nouns in *-éta-h2 (Gmc. -iþo:, Skt.
-ata:, with *o from the adjective e.g. in Slavic *-ota).

This analysis can now be used to explain the behaviour of the
causatives/iteratives in *-éje/o-. Supposing that they are denominative
*-je/o- present stems derived from Rasmussen's O-fixed thematics, the
derivation runs as follows:

*lOgH-É-jE- > *logH-é-je/o- 'cause to lie = lay'

with the first *E keeping its accent presumably because of its
contrastive value (agentive *logH-ó- 'lying' vs. *lógH-o-s 'lair,
ambush, etc.'). However, accent retraction took place in Narten roots
like *swe:p-, producing *swó:p-i-je/o- (Lat. so:pio: 'put to sleep',
OIc. sø:fa 'kill'), with the reduction of the thematic vowel operating
as expected. Also the suffix *-tó- in deverbal adjectives seems to have
drawn the accent from the first thematic vowel, also causing its reduction:

wOg^HE-tÉ- > *wog^Hitó- > Skt. va:hitá- 'caused to be conveyed'

The vocalism and accent of the Gk. okHetós 'channel, drain' may
represent a compromise between adj. *-i-tó- and the substantivised coll.
*-é-tah2, while Skt. vahita: 'flow, current' shows levelling out in the
other direction.

More details on demand. Any thoughts?

Piotr