Yeah the trick is knowing what the performance of the as-built vehicle will be before building it. You can to some extent marginalize over this by running rando

Yeah the trick is knowing what the performance of the as-built vehicle will be before building it. You can to some extent marginalize over this by running randomized trials with different parameters for your mechanisms so that you verify that you're not too sensitive to the exact setup. The control system needs to be fairly robust to such variations. But you still don't know that you've modeled all the physics correctly until it's actually flight-proven. Picking the correct degree of fidelity in a simulation model isn't super straightforward.