A backup is useful when you can restore it to the required point with the required objects. pg_dump makes a logical database backup. Cluster-level roles and tablespaces need separate treatment, such as pg_dumpall globals, when the recovery plan requires them. A copy of a running data directory isn't automatically a consistent physical backup.
Point-in-time recovery combines a suitable base backup with continuous write-ahead-log archiving. The archive must cover the needed interval. Test retention, encryption keys, recovery permissions and the restore procedure before an incident. A successful upload doesn't prove a usable restore.
This browser can't run pg_dump, stream WAL or restore a server cluster. The tasks are operational decisions using supplied fictional evidence. They check the recovery contract, not a pretend shell command. On a real service, measure the achieved recovery time and recovered data against its agreed objectives.
A database dump and cluster-global objects have different scopes.