A performance benchmark would have easily caught this though, ask the LLM to create and run benchmarks and see how those change after each new feature/major change.

I have found test coverage reports to be really helpful as well.