One thing that Git does not do is to track files, for a good reason. Git was designed primarily to track content which means it can track code movement between files and be smarter about compression. It is not limited by architecture of a file system.
ZFS is not limited to your concept of a file system either. Any change is represented by the addition of a series of one or more blocks, which can be seen as direct equivalents to blobs in git. ZFS is copy-on-write, so you can always reference any particular content change regardless of what file it was called when you made it. Compression and deduplication are handled almost the exact same way as you'd expect.
