Stability of forcibly moving dynamic rigid bodies

[rgrieder]

I'm with a development team that tries to get something like a space shooter to actually work.

We haven't yet had real physics, just doing some simple newton law integration each tick without collision detection at all.
Furthermore, we've got network support for the current design and we don't plan on changing the network concept for physics. And there lies our problem:
We need to be able to forcibly move dynamic rigid bodies around. It won't be much, merely the deviance between the linear extrapolation and the real value. But we will probably have to to it a lot (both client and server since both will have a physics engine).

Now, I don't particularly like that method because applying forces is THE way to move objects in a physical simulation as far as I can see. So, how stable would our method be? Could there be any kind of unpredictable behavior by teleporting dynamic rigid bodies?

btw: If you wish to comment on our concept, feel free, but I do know myself that it is far from perfect considering physics.