P.S. - I've also worked in shops where bugs in software can result in serious harm or loss of life. Guess what? They don't use this approach.
Which is a good seque into the code. Yes, we used static analysis tools and performance analysis tools. We needed to know exactly how long each section of code took to execute on our target processor. We also did a lot of unit testing and had extensive integration testing. We had a dedicated build manager who ran the build daily, ran the unit tests, ran the integration tests, and ran the performance tests. They sent out a daily report. This was a full decade before anyone had created what we now refer to as CI/CD tools. All told I would say there was a 4:1 testing-to-coding ratio. Our customers never experienced any issues with our system not working properly - and these were industrial users in complex manufacturing environments.
Feature management is a serial process. You add one feature at a time and thoroughly vet it before adding the next feature. The marketing department prioritized features - usually based off what customers needed for their upcoming projects. We released on a set schedule so customers could plan their updates.
If you are working on a heart pacemaker, 99% isn't good enough, but you do have to define what is. How many nines of reliability does a pacemaker need? Once you have the answer to that, you'll be able to figure out how many iterations of the 80/20 rule you're going to have to go through to get the reliability you want. Obviously in real life it's not as clean as always being 80/20, but generally the idea that the last 9 of reliability takes as much effort as all the nines that came before it holds pretty well.