Thanks for sharing, I was not aware of dar.
A couple of months ago I changed half of my backup toolchain from rsync incremental to borg, which was eye opening. The chunk-wise deduplication makes a tremendous difference and I am convinced that this is how backup should work. As I can always mount and check my backups with a fuse file system, I don’t feel the need anymore that the current state of my files should be accessible in the backup (as it was with rsync). I tested this with more than >20 TB of backup data (8 TB with rsync incremental, 2 TB deduplicated) and it worked flawless (comparing every bit between original files and backup). A big benefit of borg that it is very simple to get old backups into borg and also to get stuff out of the chunk store again.
The other half of my strategy is still git-annex (for media where chunk-wise deduplication is less important).