I would never call software good to go until I have profiled it to weed out all the places where a slow technique meets an important path.

In my experience profiling and optimization is usually sort of like a victory lap - low hanging fruit for substantial differences.