If you want to collapse just run a system at 100% for baseline, there's no slack, there's no capacity to meet new demands, you're just running a permanent failure mode if there's any perturbation in the system.
I would add it's also easy to misclassify when dealing with high-risk, low-probability events. The fact that someone hasn't witnessed something happen in their entire career can lead them to think any mitigation against that event is waste.
Except ... the collapse never happens. Once your engineers burn out or age out you just hire fresh meat and the cycle repeats. The issue I have with these types of articles (and books like Peopleware / Slack) is they never provide any actual metrics that may convince the beancounters to try a different approach.
Your systems just continue working at full capacity when all your engineers burn out and you hire new people who know nothing? Huh, that's a new one to me.
Your baseline capacity is always lower than it would be with experienced, fully trained, happy engineers. It seems normal because the system doesn’t support anything better.