Well it's more like systems show warning signs way before a total crash (like a website getting super slow before it goes down). There usually isn't one single "right then" moment. It's a bunch of small failures that add up until things just stall. You'd see errors pile up, or a vital part of the conversation just stop working right. The final click is just the last step in a longer process.
Ha! Yeah it's like hearing a weird noise in your car for weeks and acting shocked when it finally dies on the highway. The system was basically screaming for help long before it quit. Those slow response times and piling errors are it begging for a fix. But we all just keep clicking until it fully gives up. Kind of funny when you think about how we ignore all the obvious signs.
Actually, a server can hit 100% CPU usage and still take hours to fully die. The real sign is when error logs show the same failure every few seconds, or when response times go from 200ms to 2000ms. That slow climb is the actual failure happening, the final timeout is just when we notice it.