This picture from August 2006 made the headlines because it offers a new way to see the presence of dark matter in the universe (the key of course is to paint it blue). The depth of the gravitational potential deduced from gravitational lensing is marked blue, while the matter that shines ordinary photons is marked red. The picture is interpreted as showing two clusters of galaxies that have recently undergone a head-on collision. The dark matter components (and also most of the ordinary stars in the galaxies) just passed through each other with little or no interaction, while the interstellar gas made of familiar protons and electrons collided and got left behind. The observation of the Bullet cluster gave another blow to dark matter alternatives like MOND-type modified gravity theories: in the latter context it is hard to explain why the gravitational potential is not spatially correlated with the ordinary matter distribution.
What might be a little less known is that the Bullet cluster is not the only one. In August 2007 Abell 520 aka Train Wreck was revealed:
This one is much more messy. In fact, in this case the dark matter interpretation is less straightforward. The reason is that the galaxies seems to have been removed from the densest core of dark matter, and it is not clear what mechanism could have caused it. Then in August 2008 we had a pleasure to meet MACS J0025.4-1222 aka Baby Bullet:
which is another pretty clear evidence in favor of dark matter. Apparently, galactic collisions happen every year in August, so next month we should be presented with another picture of this kind :-)
The most important thing about these observations is that they look cool in pictures. But they also carry some practical consequence for particle theorists who sweat to construct models of dark matter. From the fact that the dark matter components pass through each other so easily one can derive a constraint on the self-interaction cross section of dark matter. The paper of Randall et al (not that Randall) based on the analysis of the Bullet cluster quotes the limit
$\sigma/M \leq 3 \cdot 10^3/GeV^3$.
That's an order of magnitude better than the so-called Spergel-Steinhard bound that can be deduced from the dynamics of our galaxy. While this bound is irrelevant for a standard 100 GeV WIMP, it might be a useful constraint for recently popular theories of dark matter where the dark sector consists of strongly interacting particles bound by some new unknown GeV scale forces.