Used BorgBackup 1.x with rclone, backing up in a local repository and then sending it to S3 with Rclone. For the way that borg works with files, most of the historic data lies on old, untouched files in the repository, only a catalog and the new blocks create new and update files. So it was great for using some of the S3 tiers that move files that doesn't change for some period automatically to a cheaper class.

Having both together makes easier to get this kind of use case.