> But if you send the point you are shrinking towards to infinity you get the naive estimator again.

You get close to it but strictly speaking wouldn’t it always be better than the naive estimator?