As I posted in the other thread, the negative voltage seems really counter-intuitive. There seems to be a lot of ringing going on, but that might be in the setup (sounds like you have tons of inductance and capacitance. That would also be the case in a real car, but might be significantly different, and hard to replicate. I would try to keep ground leads really short to prevent ringing and magnetic pickup.
One simple experiment would be to install a freewheeling diode and see if it changes anything, I'd just use whatever you have in hand, even 1N400x/540x. They might be able to handle the current since it's a short peak, and if you blow up a $0.01 diode, who cares? A freewheeling diode wouldn't help against negative peaks (it would only be reversed biased or breakdown), but should help against EMF that behaves properly according to textbooks
.
The peak lasts about 30ns, that's way too fast for a 50/60Hz rectifier diode, so a schottky might work. The frequency seems close to the bandwidth of the scope, so the peak might actually go even lower. I'd be careful with the duty cycle, since a ~30MHz 400V peak is likely beyond the specifications of your probe/scope (even if it's for only 100ns).
Does setting the CH1 vertical setting to 100V/div change anything in the second situation? In that case the voltage might be induced by a current into the probe (magnetic pickup by the ground lead?). Maybe the ~15pF of the scope probe over the IC shunts the spikes?