For Linux we use X11 for high quality samples (if it is available and accessible), but for Windows we rely on PsychoPy, which can be improved. Please check if there is a way to improve sampling quality on Windows, and implement it. Of course, PsychoPy sampling should always be a fallback method.
So basically there should be an object in the cfg dict (or function or function-pointer) the handles sampling, and this gets picked when the experiment is started, according to the platform and available / optimal methods.
For Linux we use X11 for high quality samples (if it is available and accessible), but for Windows we rely on PsychoPy, which can be improved. Please check if there is a way to improve sampling quality on Windows, and implement it. Of course, PsychoPy sampling should always be a fallback method.
So basically there should be an object in the cfg dict (or function or function-pointer) the handles sampling, and this gets picked when the experiment is started, according to the platform and available / optimal methods.