Here are some good explanations on the rule:
We have a highly-available cluster based on StarWind https://www.starwindsoftware.com/starwind-virtual-san and thus having 2 copies of data as a synchronous replica and a third copy as an on-site backup (which is part 3 of the rule). Obviously, cluster is running on primary internal storage and backups are stored on a separate NAS (wich is 2 part of the rule). And we have a VTL virtual machine https://azure.microsoft.com/en-us/marketplace/partners/starwind/starwindvtl/ running in Azure that hosts our offsite backups (which is part 1).
I don't agree that StarWinds provides two copies of data - this is like saying that RAID 1 is two copies of the data. They are in real time sync, so if one become corrupted, so does the other.
This is true! Unless snapshot-based async replication is configured instead of a sync. I mean 1+1 or 2+1 instead of HA(2).