Nice to hear from you again (Remember you from the old WFA days :-))
1) It is indeed agnostic and should work on any NFS source, in any location (On-premise or cloud-based).
2) Cloud Sync works on relationships (Though not SnapMirror one). You set your source NFS (Full volume or sub-folder) and target a bucket
(Or a folder within). Returning the data from S3 is another relationship, where you select the source bucket (Or folder) and target an NFS export
So yes, You can definitely return the data not to the original location.
With that said - Cloud Sync currently sync your data. If you add a file it would be synced on next schedule. Same for modify. If you delete,
we delete. We are adding a feature to allow NOT to delete on deletion.
If you are bent on archival with versioning, i'd offer you to explore our ONTAP Cloud as well:
In the upcoming 3.1 release (Slated for next month) we will support Cold HDD EBS type, which offer block storage
in price per GB lower than S3. Worth looking at.
3) You can access the target files. Both datasets are active and available at the same time. You can definitely use them without Cloud Sync.
Small note that ties into this question from the previous one - The return direction (S3 -> NFS) release date would be annouced soon and till
then you can use other means for the return journey.
Hope that's helpful
Cloud Solutions Architect Team Lead
NetApp Engineering IL