This definitely isn't the first time this has been done. Maybe it's the first time anybody has done it with an unnecessarily large cluster of 3000 (all infected) computers. I also think this study is flawed and mostly pointless. First of all, command and control-style botnets are getting easier and easier to mitigate. The real threat is from peer-to-peer botnets. The most useful research taking place as of late is not being done in a closed environment cut off from the rest of the world on a botnet that hasn't been a threat for several months. That research is being done by taking over or infiltrating known botnets that are using newer peer-to-peer botnet protocols [T. Holz, M. Steiner, F. Dahl, E. Biersack, and F. Freiling. "Measurements and Mitigation of Peer-to-Peer-based Botnets: A Case Study on Storm Worm." In USENIX Workshop on Large-Scale Exploits and Emergent Threats, 2008.] and [B. Stone-Gross, M. Cova, L. Cavallaro, B. Gilbert, M. Szydlowski, R. Kemmerer, C. Kruegel, and G. Vigna. "Your Botnet is My Botnet: Analysis of a Botnet Takeover." Technical report, University of California, May 2009.] Also, instead of infected every single computer on the cluster, they should have studied more about the ways the botnet spreads by only infecting 25% or so of the network.
Other useful projects related to peer-to-peer botnets is in trying to be one step ahead of the botnet developers. These kind of projects predict what the new peer-to-peer botnet protocols will be so they can better protect computers against being infected by them [Günther Starnberger, Christopher Kruegel, and Engin Kirda. "Overbot: A Botnet Protocol Based on Kademlia." In Proceedings of the 4th Conference on Security and Privacy in Communication Networks (SecureComm’08), pages 1–9, 2008.]
I just think this "research" project is getting more press than it should while others that are doing more aren't getting as much.