Historical realism and prenasalized stops

For those phonology buffs out there:

I've recently restarted work on Tjaren, a language I began in high school
and haven't touched for at least five years, and discovered that, purely
accidentally, I seem to have created a series of prenasalized voiced stops
(which was certainly way beyond my conscious ability at that point in my
linguistic career).

From a historical standpoint, I think it looks like there were originally
three stop types: [d] [t] [t_h], which in initial position went through the
following transformation:

*t_h > t
*t > d
*d > nd

Intervocalically, the new voiced series was lenited (to [D] in this case),
and certain rules apply to the prenasalized stops: before a stressed vowel,
they lose their nasality. Thus we have

vedái "I drink"
ávendu "I drank"

So then, the *question* is, does this seem reasonable? That is to say, have
any of you ever seen a language behave anything like this? Is it in
violation of any general rules of historical change?

In case it's useful, here's the full chart of ordered changes for the
alveolar series:

1. *d > nd
2. *t > d
3. *d > D / V_V
4. *t_h > t
5. *nd > d / _V{stressed}


Thanks,
   Josh

----------
Josh Brandt-Young <vionau@...>
"After the tempest I behold, once more, the weasel."
(Mispronunciation of Ancient Greek)

Reply