After reading this thread and a few other similar ones, I tried out BorgBackup and have been massively impressed with it's efficiency.
Data that hasn't changed, is stored under a different location, or otherwise is identical to what's already stored in the backup repository (both in the backup currently being created and all historical backups) isn't replicated. Only the information required to link that existing data to its doppelgangers is stored.
The original set of data I've got being backed up is around 270gb:
I currently have 13 backups of it.
Raw; thats 3.78tb of data.
After just compression using zlib; that's down to 1.56tb.
But the incredible bit is after de-duplication (the part described in the above paragraph), the raw data stored on disk for all 13 of those backups: 67.9gb.
I can mount any one of those 13 backups to the filesystem, or just extract any of 3.78tb of files directly from that backup repository of just 67.9gb of data.