I remember reading a paper that dealt with this issue. The setup was that a spherical mass hit another one into a wormhole that sent it back in time. The velocity was such that it would deflect the first mass before the initial collision, creating a casual paradox. If I remember correctly (and it has been years since I read the paper), they found that the second mass was garaunteed to come out of the wormhole with a velocity that would cause the first mass to hit the second in a way that would result in the new velocity, leading to a consitent system; however the final velocities were probabalistic, not deterministic, of the initial conditions.

Again, I have not seen that paper in years, so my memory might have just made half of that up.