Explore a groundbreaking approach to improving fault tolerance in replicated state machines through this 20-minute conference talk from OSDI '20. Dive into the Copilot replication protocol, the first 1-slowdown-tolerant consensus algorithm that maintains normal latency despite the slowdown of any single replica. Learn how Copilot utilizes two distinguished replicas, dependencies, deduplication, and fast takeovers to achieve superior performance in the face of slowdowns. Discover optimizations like ping-pong batching and null dependency elimination that enhance Copilot's efficiency. Compare Copilot's performance against Multi-Paxos and EPaxos, and understand how it uniquely maintains low latencies when a replica slows down. Gain insights into the protocol's design, implementation, and evaluation, making this talk essential for those interested in distributed systems, consensus algorithms, and high-availability architectures.