A/B Testing with existing release builds

Hi,

I just have a quick question before proceeding with an analytics A/B test, which is what happens if some users have access to existing versions that do not handle new remote settings?

E.g. is a user still segmented without their settings being changed, potentially skewing data? Or is a user only considered part of the A/B test once they pull a remote setting?

If it would be an issue I assume the best practice would be to only start a test when a new build is already available that all new players will be using?

Thanks for your help!