How do I turn on-call incidents into lasting reliability?
You keep the system alive when it breaks. That matters, and you are good at it. But you have seen the same fires return, and you know that surviving an incident is not the same as learning from it. That instinct is exactly right. Lasting reliability comes from treating each incident as information, not just an emergency to be survived. The fix in the moment stops the bleeding. The real gain comes afterward, in the quiet, when you ask what allowed this to happen and change that one thing so the class of problem cannot return. So after your next incident, hold a short blameless review and write one root cause and one concrete change. Just one. Track it until it ships. Do that every time and, slowly, the pager goes quiet. Reliability is not a heroic act. It is a hundred small fixes that no one sees.