> Szemerenyi, as I recall, suggests they all represent plain old h.
Szemerenyi is so anti-laryngeal as to be in danger of looking old fashioned.
His reconstruction of PIE phonology in "Intro to IE linguistics" requires no
laryngeal, but has aspirated voiceless stops, and long vowels. He does say
(p139, section 6.6.9) "Only one laryngeal is to be assumed", then for
typological reasons says "it seems to follow that the laryngeal .. was
.../h/". He then ignores the /h/ totally.
He specifically rejects a laryngeal in the *po = "drink" root (p141) and
elsewhere.
There is only place where he actual requires a laryngeal, and there he slips
it in without comment and runs a way from it. It is in his discussion of
the perfect of Skt root, where the reduplicated form shows unexpected
lengthening.
So Szemerenyi is hardly good evidence
Peter