Dark matter just strikes me as too much like phlogiston. The numbers don’t add up, so we add dark matter and bang on its properties so the numbers sort of work out. Dark matter might work with cosmological models, but it doesn’t handle galaxies all that well. That’s not a great recommendation for it. We have lots of galactic rotational data and we can collect more. There’s only one universe, so the sample size there is N=1.
If the argument is that MOND doesn’t work much beyond the galactic scale, that is reasonable. It probably doesn’t work well near the event horizon of a black hole either. That’s a common problem with physical theories. We know our physics has blind spots with regards to regions of high space-time curvature. We know there is weirdness at high accelerations. MOND postulates that our theories don’t work the same in regions of low space-time curvature either.
It’s rather obvious our theory of gravitation has problems. Everyone admits that it breaks down when quantum effects dominate. This debate really comes down to choosing a research program. Are we more likely to get somewhere extending MOND in a data rich environment or by improving the fit of dark matter models in a much sparser data environment? My sense is that dark matter is too easy, and much less likely to provide fertile ground.