Generally speaking , the anycast concept is mostly deployed with regular IP unicast forwarded services (like Google DNS), but aside from this, anycast is in fact one of the design criteria established to achieve source redundancy within a multicast domain.
The concept is quite straightforward: two or more appliances feed their multicast streams, as part of the PIM/IGP domain while also announcing the multicast source as 10.1.1.1/32, as in the example below (or a /128 in case of IPv6).
With this in place, the multicast streamers play out in the same S,G group, each one with the same source address.
Joiners will get the flow according to the shortest path, which also provides a reduced end to end latency.
Should of the multicast streams get errored ,or if the appliance breaks down, the more specific source advertisement stops and the joiners will RPF towards the source that is still active, according to the recalculated IGP topology.
This design offers an optimized source placement , making the most of anycast benefits, plus providing a robust and quick failover scenario.