Basically normal matter goes both ways, back-and-forth, compresses and expands, but dark matter goes only one-way, only expands. And this can be seen in the microwave background.
It should be considered "a nice feature" that LCDM explains this but when you have a mass distribution that can be arbitrarily assigned in space, it feels like there are a lot of "nice things you could explain".
For example, they tried to explain the precession of mercury with a hunk of dm orbiting the sun called "Vulcan" (we now know you don't need dark matter to explain this). I believe the Vulcan steps in SF are named after this hypothetical dark matter planet. It's in the same neighborhood as mars, Venus, and Saturn and built around the same time
Both DM and MOND have problems explaining various data and do well at others. DM in particular seems to require a lot of parameters and tuning to make it work.
For example, Korean scientists went looking to disprove MOND by showing there isn't something called the External Field Effect, which they found as a result of their studies.
MOND predicted that JWST would see old galaxies, which mainstream astronomers are now flipping out over.
MOND predicts that binary stars that are far apart would enter the MOND regime. Two studies confirm this observation. A third refutes it but there appears to be methodological problems in this third study.
Having been a scientist, I have little faith that the median practicing scientists have any concept of discernment in terns of following the scientific method.
You're supposed to flip your opinion on models when predictive data shows up. As far as I can tell LCDM can explain a lot of stuff but it hasn't predicted much.
Think of it this way. When you train a ML model if you have a ton of parameters you run the risk of over fitting unless you have a regularization scheme. LCDM easily has at least one parameter per galaxy, and the only regularization is "can we come up with some explanation for this?" Over fitting is a huge risk for LCDM.
By contrast, MOND has one (maybe two or three, if you believe in relativistic MOND) parameters for the whole universe.
Which is why it's making a priori predictions versus a posteriori explanations, it's very hard to make a priori predictions when your parameter for any given galaxy could be anything.
It should say something that "why don't we have a picture of it: is the center of the milky way a blob of dark matter and not a black hole" (paraphrased of ciurse) was a real publication in between the time that we "took a picture" of a black hole in one galaxy and before we had done so for Sagittarius X-1
In fact LCDM shouldn't predict the bullet cluster, a collision that fast between galaxies is extremely rare in LCDM models.
Do you know of any studies that try to combine them?
Like an eight-planet star system, or a star system with a hot Jupiter, or a multi-star system with no planets, all arising within the same giant molecular cloud (GMC)? Even though the cloud-elements behave sanely, even small arbitrary changes in initial values can produce dramatically different later arrangements. Replace the galactic GMC with a dark matter filament in the cosmic web, and why be surprised that all sorts of galaxies form within?
I don't think there is reason to fear arbitrary mass distributions so long as they obey constraints (if you're interested in the maths, the Einstein constraint equations are a good starting point you'll find in textbooks -- Robert Wald's chapter 10 is good and mathy, Alcubierre's or Baumgarte and Shapiro's numerical relativity textbooks are good alternatives for people who think computationally -- and lecture materials). We know empirical constraints and lots of solutions at many scales in the known universe already, and that constraint equations can be found for many other classical field theories; Dirac and successors (Regge et al., Rothe & Rothe et al.) are good resources wrt quantum field theories.
Roughly the slogan is to lay down some arbitrary but plausibly generic configuration and see how it evolves. Do it often enough and you should get familiar structures, or you're back to the drawing board. Large cosmological dark matter simulations that do exactly this are new hat. Standard cosmology solutions often get compared in the same project with alternative theories, sometimes following the spirit of the Standard Model Extension in parametrizing everything possible and plugging in different coefficient values, sometimes simply trying to trace cosmological implementations of an IV formalism on some published alternative to general relativity. For example, Vogelsberger et al., <https://www.nature.com/articles/s42254-019-0127-2>, <https://arxiv.org/abs/1909.07976v6> in which §5.3 has a brief literature review.