Because then you have a ticking time problem. Some automatic service might just silently start failing deep in your tech stack. And yes, you can monitor that. But then your monitoring software might fail, etc.