> If I understand it correctly, then it means if a disk is provisioned using thick eager zeroed, the CBT would consider all disk sectors is in used thus backing up the entire disk rather than just the used disk space.
That case is theory only - in normal use you first create the disk - and then sometime later - you create a backup job.
So the starting point for the CBT database usually is a populated disk.
Only this case would match: you start with a thin provisioned empty disk - then create the first full backup before you even installed an OS - and then you inflate the thin provisioned disk to the full nominal size with a comand like
dd if=/dev/zero of=/dev/sda
Now that may cause the issue you think about - though even now CBT probably assumes that non-allocated blocks in a thin vmdk can be regarded to be in the same state as the same blocks wiped with zeroes after inflation or a dd command