How many transistor shrink steps can we still reasonably expect before it becomes insane or unreliable? I wonder at which scale ion doping has to become near p

How many transistor shrink steps can we still reasonably expect before it becomes insane or unreliable? I wonder at which scale ion doping has to become near perfect in order to get the expected reliable behavior?