This makes perfect sense
. And yes, this is perfectly possible today. You can install one VBR server somewhere centrally. Let's call it the HQ for the moment. Then, at the 2 other locations (let's call them remote1 and remote2) you can install a repository and some other roles (mount server, which is by default on a repository) and even proxies depending on your needs. (You will need to draw it out because some remote locations might require some additional roles depending on the hypervisor/ storage used and so on...)
Now, the VBR server in HQ will be the command and control. So some data (commands/ metadata) will be sent over the VPN but the actual backup will remain local (if jobs are configured to the right repository and use data movers in their respective sites). For restore, same thing. VBR will be Command and Control, but the actual "work" and data movement is being done at remote1 or remote2.
While I explain it here (or at least I try, feel free to disagree
) as simple as possible, it does require some planning and putting the right components at the remote1 and remote2 site (but all components still can run on 1 single VM, again depending on size/ setup etc...)
Hope it helps