Just note.. 5 TB is a huge volume for DUplicate cmd to handle.. esp if the same network is being used by your production databases.
We once ran a duplicate restore for 1 TB data.. that ate up the entire bandwidth.. ran for hours .. and had to benally killed.. as it was impacting Production Performkance.
yes. you can. you simply need to shutdown one of the nodes (target nodes), alter the init for cluster_database=false, and then do a normal duplicate.
yes, you need an auxiliary, because you are going to be connected to both DBs: target (production) and auxiliary (test).
No problem there.
Regarding time on network. you will only use the network heavily if you run a duplicate from active database.
if you run a normal duplicate, the command uses already existing backup pieces, with minor network communication between the target and auxiliary.
RMAN only connects to get the CF information.
From a recent experience, I duplicated a 4TB database in 23hours with no noticeable burden on production. So you have a baseline.
hope this helps.
On production exadata, i use rman to backup on zfssa storage. I mount the zfssa by nfs on second (test) exadata.
on second (test) exadata, I create a sample rac database and start nomount only one instance. I start rman and connect to target (production), auxiliary (local test) and catalog.
start simple duplicate command. After 11h dupicate finished and test db works fine.