>I don't know if there is any room for supposing <sinati> was reanalyzed to
>make it fit Sanskrit grammar.
You seem to be suggesting that an original root *sin- was reanalysed as
si-n, from which other si- presents, aorists, perfects, and so on were
formed, while all original aorists and perfects and adjectives and nouns
from *sin- were lost.
The process is in itself not impossible (even if improbable), but there is
extra information: Class 9 presents all have a nasal infix in a laryngeal
root. sinati shows the usual forms of such a root, namely *sineH- (strong
forms with -ne- infix); *sinH- (weak forms with -n- infix). These appear
in the three-fold forms of the class 9 present as strong sina:-, weak sini-
(before a consonant) and sin- (before a vowel).
If the original root were *sin, these forms are difficult to explain - and
impossible within the ordinary structures of Skt. Therefore such a
suggestion seems very unlikely.
Peter