In production operations this plays out concretely. The systems that hold up under incident pressure are almost always the simple ones. Clever optimizations are the first things to behave unexpectedly at 2am.