I think a combination of:

1/ ZFS datasets with hourly (or daily) snapshots

2/ Samba with vfs_fruit

Gives the peace of mind that even when the sparsebundle shits the bed, you can rollback to a suitable snapshot and only lose a small period of backups, rather than having to lose the entire history and start again from scratch.

(I say when, not if, through considerable experience over the last 15 years that it will always, inevitably, shit the bed.)