GP doesn't say it's never happened, just that the typical programmer isn't going to kill someone with a buggy password complexity validator. By and large, the standard programmer does not hold life and death in their hands when navigating callback hell.
Again, the typical programmer doesn't kill someone when they write a bug. Judging from the backlogs of each company I've worked at, not a single PaaS, SaaS, BaaS, CaaS, DaaS, FaaS, GaaS, HaaS, JaaS, KaaS, LaaS, MaaS, NaaS, QaaS, RaaS, TaaS, VaaS, WaaS, XaaS, YaaS, ZaaS, or other would have a living customer base if one bug == one death.
There are edge cases and there are certainly plenty of times when software bugs can kill people. However, to say that the typical programmer holds life and death in their hands with every keystroke is an extreme over-exaggeration and I think you know that.
These two examples are interesting. They're both cases where what was being created was a system where software was an important component, as opposed to the software written by the vast majority of us where the hardware components of the system are always the same (monitor, keyboard, etc.) This is the same distinction in Diamond v Diehr for when software might be included in a patent. I always thought the US Supreme Court made a good decision there. Unfortunately they were later overruled by lower courts. (For legal experts out there about to correct me and say that lower courts can't overrule higher courts, I wish you were right.)