Pre-PIE had three vowels: *e, *a, and *o.
These vowels entered into combination with the 'laryngeal', *H, to produce
*e:, *a:, and *o: in PIE.
All pre-PIE short vowels became PIE *A, the Ablaut vowel/segment, which
eventually had the manifestations *e, *o, or *Ø; which form *A took was a
function of stress-accentual conditions.
PIE retained pre-PIE *e:, *a:, and *o: unchanged.
In many PIE-derived languages, the zero-grade of *e: was *e; of *a:, *a;
and of *o:, o.
Indo-Iranian changed all PIE *e, *a, and *o to <a>.
In open syllables, *o was further modified to <a:>.
Indo-Iranian changed all PIE *e:, *a:, and *o: to <a:>.
The zero-grade of Indo-Iranian <a> was <Ø> where possible;
although various devices were used to ameliorate difficult combinations.
The zero-grade of Indo-Iranian <a:> was <i>.
Indo-Iranian <a> + <y> was <e:> (/ey/); the zero grade of
<e:> was <i>.
Indo-Iranian <a> + <w> was <o:> (/ou/); the zero grade of
<o:> was <u>.
These proves the vowel + glide nature of these two Inod-Iranian long
vowels.
If Indo-Iranian <a:> was /aa/, the zero-grade should have been /a/,
i.e. <*a>; but it is <i>.
Therefore, Indo-Iranian <a:> _cannot_ have phonetically been
/aa/.
To judge by known Indo-Iranian processes, the phonetic realization of
<a:> would have had to have been /ay/, the zero-grade of which would
unforcedly have been /i/, <i>.
If pre-PIE *H were phonetically realized as /ç/ in PIE, for Indo-Iranian, a
change to /y/ would involve simple voicing.
I have no doubt that at the time of Panini, <a:> was pronounced as
/aa/. A development of /ay/ to /aa/ is very common. But because of its
zero-grade realization, /i/, <a:> cannot have been /aa/.
To summarize, the crux of this problem is understanding the correct
phonetic nature of Indo-Iranian <a:>.
Some of the readers who are not native English speakers will question
whether long vowels are double-length by definition. In English, _all_ long
vowels involve glides.
Patrick