Dark matter is pretty weird stuff. We've seen tons of evidence for it: the way galaxies behave in clusters, the way individual galaxies rotate, and just in the past year two monster pieces of evidence supporting it were found (the Bullet Cluster (#4 on that list) and the COSMOS survey).
The deal is, dark matter doesn't interact with normal matter except through gravity. Two clouds of DM could pass right through each other and not collide like normal matter, but their mutual gravity might distort the shapes of the clouds. Dark matter also appears to hang out in giant clouds surrounding galaxies, and in theory may be the seed which helped cause galaxies and clusters to form in the first place.
The evidence for this has been really good... but then a monkey got thrown in the wrench.
The image above was just released, and shows an optical image of galaxy cluster Abell 520 with an X-ray image from the orbiting Chandra X-ray observatory overlaid on it. The optical image shows stars in our Milky Way and galaxies belonging to the cluster. The red fuzz is hot gas made of normal matter, heated to millions of degrees in the cluster. The blue fuzz is the location of the dark matter in the cluster, inferred by its effect on the light of background galaxies (I describe how this is done in detail here).
The problem is, where there appears to be the most dark matter, near the center of the cluster, there are very few galaxies. And there are areas where there are lots of galaxies, but little dark matter. This is the opposite of what the theory predicts! In general, the DM and the galaxies should stick together, so wherever you see one you should find the other.
Note that I said "in general". While astronomers are scratching their heads over this, I would caution that we are only just now starting to be able to make maps like these, and it doesn't surprise me that all kinds of weird things will come up. While the Bullet Cluster was a brilliant confirmation of DM theory, as was the COSMOS survey, not all galaxy clusters are equal. Since we have seen a whole pile of evidence supporting DM, it looks like Abell 520 will give us insight into details of the theory (as opposed to overthrowing the idea of DM altogether).
Do DM particle interact more than we thought? Could the galaxies in Abell 520 have interacted in some way as to separate them from the dark matter? Is there some other aspect of the theory we missed, or are wrong about? The astronomers who took this data will be getting more time not only on Chandra but on Hubble as well to investigate this mystery.
The beauty of this, as usual, is that by getting contradicted over what we expect, we learn more about what the Universe is trying to tell us. Confirmation of theories is always good, of course, but it's that occasional slap in the face where the real progress is made.