I think I've undertood what you are asking, and supposing the two sites have proper routing and dns in place, I have a feeling it may work already today. Let me explain why: during an activity, data is sent over the wire using source repositories (you said cloud connect repositories, so I believe you are talking about backup copy jobs) towards cloud gateways. At the beginning of each backup copy, VBR refreshes the list of cloud gateways, and passes it to repositories (or gateways for the shares and dedupe machines), which then contact directly the cloud gateways using either public ip or dns name. Regardless which one are used, if the vpn between the two sites is configure to route only the subnets of the other site, everything else should go over the local internet connection, and so this would happen also to the backup traffic.
I have some more doubts for the partial failover, but again unless you have a stretched network, the two sites have two different subnets, so Veeam would have deployed at least two NEA, one per subnet. But seems you are talking about backups, not replication.
Hope I've been clear enough.
EMEA Cloud Architect @ Veeam Software
Veeam VMCE #1