A few years ago, I was working on a backend service that looked completely healthy. Tests were green. CPU usage was normal. Memory looked fine. No obvious exceptions were showing up in the logs. And yet, every few hours, requests would suddenly slow down. Not fail. Just become painfully slow. The API normally responded in around 80–120 ms. During the problem, some requests were taking 8–12 seconds. Then everything would recover by itself. Those are the bugs I dislike the mos...