When requests start piling up, I look at where they spend their time and how much work the service takes on at once. The same checks apply when AI wrote the code.
What happens if the service restarts halfway through a request? Questions like this help me decide what to ask AI to build and what to test before release.