Many senders to one receiver is sort of the opposite of an expected multicast design. So, first, do you really, really need the camera to use multicast to send to a single receiver? (From what you describe, the same ten cameras, of forty, send to the same server, of four, correct? Or, do all cameras send to the same multicast address which is also received by all four severs, which in turn, determine which ten cameras, of your forty, that they want to process?)
If multicast is the only option, you're going, especially in this case, to want to use sparse mode.
Since you'll be sending traffic over VPN tunnels (in this case DMVPN), you'll want to insure your routers don't need to fragment the video packets. You may need to determine if the cameras are generating packets larger than the tunnels can handle, w/o fragmentation. If they are, then you'll want to see if you can either use PMTUD, to respond to a "packet needs fragmentation" message from the router (BTW, I recall [?] router may also need to be configured to so notify the sender) or not generate too large packets to begin with.
CCTV cameras, depending on video codec, and what the camera "sees" might vary much in the bandwidth used. Does you hub have sufficient bandwidth to support all forty remote sites, sending heavy streams, concurrently? Will any other traffic be using this topology beyond your CCTV streams? If so, again, does your hub have sufficient bandwidth?
