Does anyone else see a problem with a session based approach to a/b testing (that Google Optimize uses)?