Thank you for reaching NetApp community.
Yes what you understood is right
In all of the scenarios (Snapshot, Backup policies) the blocks comprising a file's contents get moved out to the external capacity tier, but the file's WAFL metadata remains in the performance tier. Files with partially or fully tiered blocks still appear to be in their original locations in the primary tier, so users can still browse directories and list files in the primary tier without any data transfer from the object storage, or even any knowledge that the file's blocks have been moved to the external capacity tier. If a tiered file's contents are read, FabricPool seamlessly reads the blocks back from object storage into the primary tier, with the only visible impact to the end user being a short delay as FabricPool transfers the blocks back. Once the user is finished with the files, FabricPool's algorithm for cooling blocks may eventually force the associated blocks back out to the external capacity tier, depending on the volume's assigned tiering policy.
Please refer to FabricPool best practices for more information
*** Edit ***
The backup policy can only be set on data protection (DP) volumes (destination targets). All data blocks (does not include WAFL metadata) in the volume are immediately moved to the capacity tier until read and made hot. When read, hot blocks are moved to the performance tier. Unlike volumes using the Snapshot-only tiering policy, there is no need to wait 48 hours for blocks in a volume using the backup tiering policy to become cold.
If the volume’s data protection relationship is broken, the destination volume will no longer be a
DP volume and the tiering policy will revert to the default, Snapshot-only, policy