A company like Paradox should certainly be able afford testers who run the game on a variety of configurations to see if optimization is necessary.

One thing I would say and this is a broad statement - generally you don't do optimization unless you know you need it. And you only do it after the thing you're writing is working correctly non-optimally. Optimize too soon, or when you don't need to just makes code an unmaintainable mess. That doesn't doesn't preclude writing efficient code in the first place but efficient is not the same thing as optimal.