I live fairly low on the totem pole, and I don't have a high-level answer to your questions, but I have some observations that you may find interesting. I am a physicist working at National Institute of Standards and Technology, a government research institute; we have been using Linux since 1994, and Open Source (OS) since forever; we had a feature article in (and our mug shots on the cover of) the Linux Journal few years back, about our use of Open Source in our laboratory. We were also honored with an Outstanding Achievement in implementing mission critical projects using OS, given by a local technology council. I am also the organizer of the local Linux User Group.

Overall, NIST computer environment is dominated by You-Know-Who, but there is a strong Linux presence, both due to a strong Unix tradition (NIST=POSIX), and to the effects of the PC/Linux revolution. A vocal group, of which I am a member, tries to get the IT management here to consider officially supporting OS , both because of existing deployment, and because we believe that it might be a good and cost-effective solution to managing the ever-increasing PC herd.

We had some encouraging meetings with management, and haven't really seen any pushback. The dominant negative reaction was probably scepticism, mostly coming from our peers, based on the "let's just buy a solution". We were unsuccessful in making a principled argument a' la' Nunez (the Peruvian politician), who proposed that governments should use OS for fundamental reasons of transparency (if you haven't read his letter, I recommend it)---perhaps because we aren't good at that kind of thing, perhaps because the current climate is not conducive to deep ideological reasoning about the functioning of the government.

We seemed to get most success by making narrow arguments about cost, security and administration. We are proposing a pilot project, using/supporting OS for internal functions, to start this fall.