[ODE] "overconstraining" in kinematic loops?

[Graham Fyffe]

Hi!  I'm still having an odd problem with kinematic loops.  I'm
finding that if I have several joints in the loop that would seem to
eliminate the same degrees of freedom, for example several hinges all
on the same axis, then the system becomes "overconstrained", i.e.
degrees of freedom are lost that should not be lost.  In one instance
I have a loop that could not rotate on any other axis except those of
its hinges, even though the loop was floating out in space!

A possibly related problem I'm having, again with loops, and again
with weird physically impossible motion, is that it seems if I push
against a joint that has limits, even where the push is originating
from inside the system, then it can cause the entire system to rotate.
 The specific example I have is a "rack and pinion" set-up I have for
car steering.  I have a rack that moves side to side on a slider,
which pushes against the front wheels via tie rods connected with ball
joints.  If I push too far with the rack, the entire car lifts up off
the ground (!)  This is not what one would expect, since the forces
come from the slider internally :P

Is there any way to fix this or work around this?  I'd be happy to
provide more detail.