Dang. No.
Don't think you can use IGAR either - so, every network region can get a DID for Inter Gateway Alternate Routing and Dial Plan Transparency.
DTP = in WAN fail, no signaling to CM. extn 1111 in NR 1 calls 2222 in NR 2, the LSP handling NR 1 knows NR2 is not reachable, so it grabs a PRI trunk, calls the DID for NR2's DPT/IGAR and passes the digits thru like an automagic auto attendant.
IGAR is the same but implies signaling is in place. But, your network-map/network-regions won't know or accommodate that core 10.x.x.x to site 1 at 172.16.x.x via hops 192.168.1.x core normal WAN is better than thru hops 192.168.2.x on backup VPN. unless those phones had different IPs entirely, CM wouldn't be smart enough to know you're on backup VPN
You do have the conditional in vectoring "if gateway 12 isn't registered then..." which could be leveraged "if no agents in skill 12 and no gateway registered in skill 12, well then...let's vector according to WAN failure!"
Now, if you had that condition, you could route to a DID at site 12 that hits the same vector/vdn/queuing. It'd make sense if site 12 has a PRI and the only agents that ever answer skill 12 are at site 12 and you have central trunks to reach them, that might be a clever reroute.
Depending on how you administer the failover timers, you might be able to get the gateway to failover to LSP but stall the timeout on the phones until the backup VPN is up, so they'd eventually come back to the main, but you'd have no way of getting calls to them if they're only tie to the core is on sketchy WAN and they can't take calls from the gateway on site cause they didn't failover with it.