Sanskrit t & d

[Peter &/or Graham]

Miguel's question (why some Sanskrit forms are quoted with an etymologically
correct final -d when all occurrences of the word in Skt would neutralise
the difference between final t and d) still intrigues me.  Is it possible
that it is a question of internal sandhi, rather than external?  The rules
are different, and at least one book says that the difference between voiced
and voiceless consonants is maintained in internal sandhi.   It then offers
the unhelpful example of tad-apas "accustomed to that work".    If compounds
of a root such as vrt maintain the /t/ in that position, there would be the
answer to the question.   Unfortunately a quick troll through my Sanskrit
literature yielded nothing - unsurprisingly.   Nonetheless, the vrt root
does not alter t to d when a voiced suffix is added - nivartitum "to turn
back".

Any reactions?  And does someone have more detail than I could find in the
library?

Peter