etymol. final r, s / optionality of ext. sandhi?

[Jacob Baltuch]

What follows is elementary. If you can't stand this kind of
stuff delete this post now.
------------------------------------------------------------------------

I have two questions which found no answer (or no explicit
answer) in either Renou's grammar or the two textbooks I've
searched (Coulson's TYS and Ashok Aklujkar's SEL). Maybe you
can help.

1. Renou gives, in flexional forms, always final r or s, never
visarga. In particular he keeps the distinction between final V+s
and V+r (V not a or aa). For example he gives for the aorist par.
of kR 2sg "akaarSiis", 3pl "akaarSur" while for the same forms SEL
would give "akaarSiiH" and "akaarSuH". Is there _any_ reason what-
soever (not only in such forms but in any hidden corner of the
language) to keep this distinction (between final Vr and Vs, V not
a/aa) or is this just a little etymological pedantry?

2. I've always wondered about the "optionality" of external sandhi
in spoken Sanskrit. That's what most textbooks state. And I've always
had trouble taking it seriously (by analogy to French connected speech
since I've only dealt with written Sanskrit) Is there really such
a thing as fully non-sandhied speech which really sounds natural?
To put it differently, what is the impression you get as an accomplished
spoken Sanskrit practicioner from hearing someone using fully non-
sandhied speech? Does that mark that person immediately as an incompetent
speaker or maybe as a somewhat slow-witted individual or is it really
just a choice?

Renou, contrary to custom, never says external sandhi is optional, instead
he says it is mandatory in general and enumerates those cases where it is
not done, but then his concern is with the attested written usage and not
with spoken Sanskrit. But now if the written record does not seem to attest
that external sandhi is optional, where does that idea come from?