I have no idea why you're talking about dial peers.
If you build a route list that contains two trunks, say to two separate border elements, and assuming the trunk is up but the far end item can't process the call, the UCM will use the next item in the route list based on the signaling received from whatever is at the other end, and a parameter that tells the UCM to either continue hunting or not depending on cause. I believe there are two CallManager service parameters for stop routing on busy, or unallocated. There should also be timeouts in the SIP profile which could mark the transaction failed.
You could have two TDM gateways or CUBE or whatever and stack them in this manner for resiliency. The concept of overflow would depend on the signaling and how the UCM is set to handle it. And, as well, if you're trying to consider how that far end device is going to play along in this scenario, it also depends what is happening or what it does. If you are using a SIP trunk out from there, let's say, and the carrier is going to intercept the call or do something weird when you hit a capacity limit, you may have to figure out how to "translate" that into something that causes routing to proceed via the UCM rather than being terminated by that device.
