This is not unusual. There's many predictions made by MOND ahead of the fact that are borne out by reality. But don't worry - dark matter will be "retrofitted" so that it fits the facts and everybody will suffer collective memory loss again.
Edit: see for example table 1 on page 12 here: https://arxiv.org/pdf/2110.06936.pdf from a review of prior expectations by both MOND and dark matter vs how they turned out against reality for a large variety of astrophysical scenarios.
MOND is not compatible with some observations. It and dark will probably be tweaked until both agree with observations. At that point they will be the same.
That's a false equivalence. MOND has made several predictions before they were observed (external field effect, linear Tully fisher relationship, early galaxies). I think LCDM's made fewer if any predictions.
Dark matter is the umbrella term for the observed discrepancies. MOND is one possible possible explanation for DM. The paper you linked doesn't compare MOND to DM, because that doesn't make any sense. It compares MOND to ΛCDM which is a competing explanation.
Am I understanding that paper correctly that their suggested solution to explain the Bullet Cluster using MOND is too introduce an additional kind of matter which can not be detected on earth and doesn't interact with light (namely sterile neutrinos)?
"dark matter but not as much" is still an improvement if you replace it with something better. If MOND predicted these observations and LCDM didn't it's reasonable to say that it is better.
Yeah that's fair. I guess it could also explain why DM has been so hard to pin down, if both MOND and DM are true then there might be DM candidates that have been unfairly ruled out.
I do think it significantly hurts the (more philosophical) argument that MOND is simpler or has fewer parameters than DM though.
So the ongoing problem with LCDM is that the parameter space for allowable particles keeps getting pushed back. We keep falsifying classes of potential particles that could be WIMPs. MOND+ some standard model-adjacent particle (like say right handed neutrinos), if we can have a phenomenological estimate of how much of them should exist, would at least be credible on account of having a constrainable set of parameters, whereas the density of WIMPs in normal LCDM has no constraint besides the same measure that we have to infer its existence (gravitation)
Edit: see for example table 1 on page 12 here: https://arxiv.org/pdf/2110.06936.pdf from a review of prior expectations by both MOND and dark matter vs how they turned out against reality for a large variety of astrophysical scenarios.