Dear Madhav,
The way I see it, the difference is simply carried over from the base form, represented by the nom./acc. (<nava> vs. <as.t.a:>). The comparative evidence suggests that the numerals from "5" to "9" were indeclinable in PIE, whatever their etymology. The inflected forms in Sanskrit are relatively late innovations (so, incidentally, are the forms of <das'a>). The case forms of "8" are torn between two analogical attractors: the inherited preconsonantal nom./acc. (the long-vowelled forms) and the analogy of the "7, ..., 9, 10" series with Skt. -a (the short-vowelled forms).
Piotr
----- Original Message -----
From: <mmdesh@...>
To: <cybalist@yahoogroups.com>
Sent: Thursday, December 26, 2002 11:54 AM
Subject: Re: [tied] Sanskrit form a.s.tau
> The dual -au frequently alternates with -a: in Vedic. However,
> with a.s.ta, the lengthening is also seen in other case forms
> such as a.s.taabhi.h, a.s.taabhya.h. This contrasts with
> navabhi.h and navabhya.h. So there is some inherited different
> behavior in the forms of a.s.ta.