So, Postgres switched, but it took them longer than the slightly larger Linux kernel project. I'm not sure that is something to be proud of.
Beyond CVS conversion oddities (sure, those suck), their install seems to do without the most powerful features of git - no merges, no author/submitter history, etc. In fact, it appears they are using it as a centralized and linearized CVS replacement, for which they might as well have used SVN (except for storing full local history).
In particular the "no merges" bit is really weird. For review, reviewing the history of a feature/bugfix in an offered pull tree would appear to be much better. I could understand an "I'll not merge if there are conflicts to resolve" policy, but that?
Yes, of course, it takes a while for people to adjust, but I think this is self-inflicted to a degree; a number of the problems (short of the conversion madness, but that's a one-time effort) seem to stem from the desire to not fully adopt git. (Which explains why specialized documentation for the project is needed, etc.)