A snapshot is not a backup
A snapshot points at the same disks as the running virtual machine, so it disappears with the datastore.
A snapshot records the difference from a base disk. It is useful before a patch, and it is still stored with the virtual machine. If the datastore, the array, or the host volume fails, the snapshot fails with it.
A backup is a separate copy on other storage, ideally in another failure domain. Keep the snapshot only for the short window of a change, then commit or delete it so the chain does not grow without limit.
Treat the snapshot as a convenience and the backup as the recovery copy. Write down which jobs create the independent copy and where that copy lives.