While working for Motorola, I worked on electronic throttle control (ETC). We spent a ton of time working to make the system "fail safe". I think we all had in the back of our minds that it was only a mater of time before we would have to testify as to our engineering decisions.

My little part of ETC involved adding a sub processor which watch-dogged the main micro. The little micro asked a series of questions of the main micro. Both processors would need to agree on all the inputs and output of the system. The little micro would also ask question regarding real time OS (RTOS) of the main micro. The main micro would need to have tasks executing in the right order to satisfy the small micro. Lastly, the small micro would ask the main micro to perform math operations to verify accuracy. Oh, and the main micro was continuously checksumming it's memory too.

Both micros had a direct hardware disable path to the H-bridge which was delivering power to the throttle plate. The throttle plate was spring loaded, so, with power cut, the throttle plate would snap to an idle position.

Next came the electro / magnetic compatibility testing (EMC). We spent months inside huge chambers testing both radiation and susceptibility. One of the tests for susceptibility involved using a zap gun to spark a 20kV spark on each pin of our ECU. Not satisfied with that, our customer opened one of our modules and used a sparking spark plug to slowly zap our board to failure. Bottom line, that throttle plate better never stick one way, or the other.

In the end, it always amazed me that the whole thing would work at all. Seemed to me that the system was always seconds away from going into some kind of fail safe mode.

No, a stray bit flip is not going to facilitate a run away car. Least not on my system!