Your question is Covariate Shift Evaluation. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
What are the primary differences between training and testing distributions (covariate shift), and how do you address this during model evaluation?