>it is
>an open question whether *H2y- yields /h-/ because *H2 is voiceless,
>or /z-/ because *H2 is also a somewhat strong sound.
I have a note from somewhere that "there was early coalescence of *(-)h2y-
into a single segment, with resyllabification:
*kreuh2-yo- > *kreu.ço- > * kreu.yo- (cf Skt kravya)
Thus always in initial position, which is why Greek has prothetic vowels
before *w- but never before *y-."