I think that the definition still works - you just need to go farther down the scale towards pure engineering, and you start running into things like anti-lock-brake sensor programming, bridge design, refinery design, etc. The constraint "no failure whatsoever" is a pretty hard constraint, and while it may limit a lot of options and take more time and care, it would let you pursue perfection and responsibility to a degree not possible with regular business apps.
Yeah, I had a boss who liked to bitch about how NASA can do all this great stuff that works the very first time. I told him we could too if we had millions of dollars, and accepted the fact that it would take a long time to deliver even relatively simple components. And of course nothing would touch the running system without weeks of QA.