Best Practice: Dataset setup going forward (encryption, root datasets etc)

Running Dragonfish-24.04.2.5

When I setup my pool “tank” back on Core I established it as encrypted with an unencrypted TimeMachine dataset present. However on Scale this now means I cannot have an unencrypted dataset to use as a TimeMachine target so that each client can encrypt its own backup and not have double encryption take place.
I also have the issue that I cannot easily replicate my data as there are multiple datasets under the tank initial one that cannot be replicated tank → dozer for example but must be tank\A to Dozer\A etc.

Going forward would it be better on new pools to establish top-level zroot encrypted dataset and zroot1 unencrypted dataset on an unencrypted pool? This way there are only two datasets to access for replication and I have the choice of encrypted or unencrypted parents.

Is there a disadvantage in performance, security or anything else to having this?

Does double encryption of data even matter?

I think I at least need to have a zroot dataset under the initial tank pool creation to avoid my current multi-source to multi-target replications.