I agree with your first sentence but I don't think that argument really supports it. Performance-based testing is not inherently flawed. You just need multiple samples. Bad sleep and stimulants can screw up a lot of other measurements, too, like blood sugar and pressure.