A Harbor design places compute in more than one failure domain and keeps a second database copy. Its diagram looks redundant, but both application paths depend on one identity service and one untested archive key. Redundancy only helps when the surviving path can complete the required business action with acceptable data.
Work from the recovery point and time objectives, then identify capacity, dependencies, permissions and routing at the recovery location. No exercise here creates real replication or measures a drill. The choices specify the tests that would support a production claim. Be precise about the difference between having a backup, restoring useful data and recovering the complete customer operation.
Test the complete surviving path against the stated recovery objectives.