being bolted to a ton of body shell automatically limits the peak operating g the unit can undergo!
Yes, it wouldn't be about peak G's, it would be about hitting resonant frequencies of free standing packages from broadband vibration.
I've seen free standing TO-220 packages vibrate clean off from just wheeling up and down a production floor for a few weeks on a metal trolley.
But that is exactly the point, there isn't any! (well very little) because the mass of the body shell damps it so effectively. Chances are, those SIL leaded components have resonant frequencies in the hundreds of Hz range. Now calculate the power required to say vibrate even 20kg of steel body shell at just 100Hz. It's massive!
Ok, mistakes have been made, people have bolted control units to flat panels that pant or suffer a critical oscillation from drivetrain driven resonance, but generally speaking it doesn't happen as modern car are so well designed/tested (not to mention the NVH aspect of having parts of your bodyinwhite vibrate at 100hz). The primary ride frequency of a typical passenger car is in the order of 1 or 2Hz, and most of the accoustic noise is extremely low amplitude at say >1000Hz, that noise won't even make it through the pcb, let alone snap components off it!
(In fact, the only time i can actually remember an issue with a control unit failing was on a rally car, where the control unit had been mounted on rubber, vibration absorbing "bobins" in a misguided attempt to reduce transmitted vibrations. What actually happened is that those rubber mounts, in conjunction with the mass of the control unit forms a critically tuned spring/mass oscillator at around 10Hz, which for a competition car IS in the normal ride frequency region. As a result, there were some failures of some large through hole capacitors on a few units. When we removed the rubber compliance, bolted the control unit directly to the bodyinwhite, all was well! ;-)