https://archive.nanog.org/meetings/nanog42/presentations/Weissner_SSO.pdf

The strict-priority queue should be limited to approximately 33 Mbps on a 100 Mbps Internet connection.
Cisco QoS design guidance warns that strict priority traffic must be bounded, because excessive priority traffic can starve other queues and cause poor application performance for non-real-time traffic. For voice and interactive video, priority queuing is appropriate, but the priority class should normally be constrained to about one-third of the link capacity in a converged enterprise design. That makes 33 Mbps the best choice for a 100 Mbps link. A 75 Mbps strict-priority allocation is far too high and would leave inadequate bandwidth for routing, signaling, business data, and default traffic. A 50 Mbps allocation still gives real-time media excessive control of the interface. A 25 Mbps allocation may work in some environments but is not the standard upper-limit design answer for strict priority in this Cisco context. The architect should also classify TelePresence accurately, mark traffic consistently, and police the priority queue to protect the WAN.
Reference topics: LLQ, CBWFQ, strict priority limits, TelePresence QoS, WAN congestion management.

Explanation
Reverse path forwarding (RPF) mechanisms are used to prevent loops and duplicate packets in multicast deployments [1]. The RPF check verifies that all multicast traffic received on a router is sourced from the expected direction, ensuring that multicast traffic is not looped back in the network. The RPF check works by comparing the source IP address of a multicast packet with the routing table, and only forwarding it if it matches the expected entry.