Since our brains have dedicated hardware for processing and regulating emotion and thought, there could at some point be the possibility of brain scans to demonstrate what areas of the brain light up in response to certain stimuli that could prove certain kinds of internal preferences for cognitive processing. As far as I know this is not used in relationship to MBTI. I don't think MBTI correlates its poles according to brain structure.
Basically I would say that as long as the implications of being some type are not being measured, then it doesn't matter whether or not you can measure a preference. If having a label that indicates a preference impacts employment opportunities and other measured outcomes, then there sure as hell better be an equally valid way of measuring that preference.