I finally figured out why machines in the cluster couldn't reconnect -- I had munged the permissions of the directory where they wanted to store their connection info. I fixed that, and they can happily reconnect after rebooting. After I figured that out, I sent a broadcast to the department and invited anyone else who wants to run a lot of BLAST queries to contact me about getting the databases they would need installed. E showed up in my office 10 minutes later -- it was exactly what her lab needed. I got the databases installed on one client this morning and added them into the radmind image for the lab. They've been running a few test queries this afternoon, but tomorrow they should be able to start using the cluster to do some real work. It's very satisfying.
I was interested to read a variety of the commentary associated with this 
boingboing posting about the so-called "rank and yank" evaluation systems. It's the sort of thing that seems seductive about all forms of human measurement, until you factor in the sociological factors: e.g. the psychological effects of inhumanely ranking people, the incentives for cheating, the potential rewards for undercutting the other people around you, etc. I'm glad I've never had to be in an environment where such a scheme has been imposed. I've always seen myself as a kind of catalyst or force-multiplier in the roles that I've had and I suspect that these kinds of schemes would evaluate those kinds of contributions fairly poorly: the person who does not directly produce much, but who enables other people to be much more productive.
StevenBrewer