Wow, as I actually flesh out some of my ideas (some with the help of Autocad) - you actually have a pretty good problem on your hands!
Even if you keep the sprocket's relative angles the same, they shift into different planes like this:
_ - <-- not to scale...
But like you have said, with minimizing the motion of the suspension; then just a little bit of self-correcting measure would make a large % difference. I have an idea, but I'm researching to find out if it exists! It would be something simple, but I don't if it would work in the real world.
Something similar to a spherical bearing to allow this motion:
But the bearing and it's housing would have intermeshing teeth or ridges to allow that changing angle but transfer torque from the housing to the rod... ehh, I don't think it exists. Bummer.