Since mbti essentially neuters it's own usefulness in the "round those percentages down to binary" stage, you could end up with a situation where an INTP and an ESFJ have only an 8% total difference in mbti-measured personality between them.

The question isn't whether people who are in the middle of each criteria exist and could easily swap classifications to their polar opposite. The question is what percentage of people would do that and whether it's high enough that there's too little signal to noise to extract any meaningful information out of the classification system.

Don't invert the burden of proof. It's up to the proponents of MBTI to demonstrate its effectiveness.

As for signal-to-noise, mbti is a useful tool for measuring the traits it measures, it just, for some reason, cuts the signal in half by removing the percentage from the label.

Without the percentages, you can never assume a persons traits don't hover at that middle ground. And thus, can never assume any part of their label is accurate.

If the occurrence of these hover cases were more like <4%, I'd be more inclined to agree with its usefulness, but as it is right now, the 4bit label, on its own, is just noise.

i'm pretty sure someone who has gotten INTP every single test in his life will behave differently on average than someone who has always gotten ESFJ

I think its a mildly useful categorization which allows a simpler and more coarse view of personality than the full spectrum of data would normally allow.