Unless you use the -force delete flag it's not immediately deleted. So that statement isn't entirely accurate.
Where is the procedure to validate the keys are no longer being stored in Cache?
There is this command:
::*>security key-manager query -node <node name>
But is that querying the exact database or location of the key that the node is using for encrypted read/write operations?
None of this explained in any detail. I also still don't understand where the data buffer comes from that doesn't allow me to stretch volumes out to the total available size of the aggregate. This test shouldn't be difficult and really easy to prove except the avaialble space calculations don't seem consistent.
I can show the available space on the aggregate, create an 8TB volume, show the available space on the aggregate. Then Delete the volume and show the available space on the aggregate and the number is significantly different than when I did the first show command. That alone makes me think there must be residual data somewhere.
The volume and key are put in the recovery queue as an accidental deletion feature. If you need it gone at that very moment, yes run -force. Or you can mod your vserver/SVM to have 0 retention. The keys are stored within the wafl meta-data which isn't accessible to the user and is deleted along with the volume.
Starting in 9.4 you can also have the ability to "shred" a single file if needed. NVE secure purge is GDPR compliant.
Couple other points:
This command will give you the side in kb count of the aggr. "run -node <node> -command df -A"
Using thin volumes you can go way over the size of the actual aggr.
Keep in mind there is ways time to clean up the data that is deleted.
There's always WAFL metadata (which includes more than keys) hanging around an aggr, so you'll never see 0.