No.
Essentially once you execute the procedure to reconfigure a secondary it diverges. Things can get… messy after that if precautions are not taken. The best way to recover is to then add the nodes from the primary site back as new members and sync from the node in the secondary site.
I like a 5 node cluster for this configuration with one of those nodes being an arbiter in a third datacenter and using member priority to keen the primary in the main site during normal operation. An arbiter can be modest in specs compared to the data nodes. This allows for network partitioning to occur, primary move to the secondary site, and recovery back to the main afterwards(given an appropriately sized oplog).automatically.