Because cgit is a bunch of binaries that expect to call each other. That makes it harder to abstract out the storage layer, and we don't use vanilla repositories sitting on a filesystem. Things are backed by some other storage abstraction, which isn't always very posix-filesystem like.
There was a google talk on this posted to HN recently, but I can't find it. In it, one of the directors of the build / testing / code review system at google was talking about how they get things working at scale. Since everyone works out of the HEAD of one Perforce repo, they end up using the map-reduce infrastructure to perform tests in the cloud for each checkout. In line with this, there are too many files, that update too often for every developer to be checking out of the repo, so they use a custom FUSE filesystem to lazily give access to files only when they're needed.
