I suggested this a few years ago and I was inspired by TEREP (predecessor of 1NSANE). It was able to simulate 5 cars with tens beams each and do 320x200 sw rendering on something like Pentium I CPU. RoR's simulation run at 2000!!! FPS. I doubt that lack of FLOPs is an issue. Even graphical representation should be feasible, it's not much different from skinning.
What I think would be real issues about simple mass-spring structure:
It affects the whole physics too much, requires massive changes.
- Center of gravity is distributed to nodes -> complicates making cars
- Realistic moment of inertia is difficult to achieve for the same reason
- Desired chassis stiffness hard to achieve as well
My idea of workaround:
Each suspension mounting point gets a 4D vector. V1, v2, v3 are coordinates that define direction in which the mounting point moves under load. Vector length (or v5) can represent spring constant and v4 damping.
Maybe even better idea: What has already been suggested, the front/rear end rotating. Rotation limited to axles. Visualization possible.