"Debunked" is probably overstating it, but Myers-Briggs has been shown to have fairly significant statistical deficiencies and its predictive claims are largely pseudoscientific. It's not complete garbage, but there are personality trait models like Big Five and HEXACO that are more robust and have more research behind them, but even those are not as strong predictors as pop management literature makes them out to be.

I never considered it to have any predictive meaning at all, so saying it doesn't seems to miss the whole point.

Predicting others' behavior on the basis of a generic quiz is a pretty silly expectation.

I suppose the issue with centrality is that not only can it indicate that you are both good at being an extrovert and good at being an introvert (when appropriate), but it can also indicate that you are bad at being extroverted and bad at being introverted when arguably called for by context.

I think what MBTI lacks is an appreciation of context - sometimes it's good to be Thinking, sometimes it is good to be Feeling. What individuals need to learn is 1. when each is appropriate and 2. if they're bad at doing one or the other, getting better at it.