As I understand it, the real dubious thing about MOND, and what makes most physicists discount it, is that it looks a lot like an exercise in curve-fitting[0]. Maybe that'll actually lead to an explanation of reality, but in general physicists are like programmers: they like elegance. They want to start with a simple equation or modification and, after applying all the math, find that it suddenly explains a bunch of data that never fit the old model and makes testable predictions.

[0] IIRC this wasn't really true of the initial form that only explained galactic rotation anomalies, but every other anomaly we credit Dark Matter for has required more and more tweaking to explain with MOND.