Persistence/Failover Problem with Pool Member

You need to force offline the low priority pool members, but rather just remove the persistence entries of it.

tmsh delete ltm persistence persist-records pool  node-addr  node-port 

As per articles & other previous Q&A's.

Note: When the BIG-IP system activates a lower priority group and its pool members are processing connections, active and persisted connections continue to use the lower priority group pool members after a higher priority group becomes available. Connections to the lower priority group pool members continue until the connection has completed or the configured persistence timeout is reached.

But you may achieve this with some Irules (Hoping, never tried)