Design High-Throughput Caching System

Design a high-throughput caching system that can serve millions of read/write operations per second with sub-millisecond latency, high availability, and strong consistency guarantees. The system should support multiple eviction policies (LRU, LFU, TTL), handle hot keys and thundering-herd prevention, and remain available during node failures or network partitions. You must distribute data across cache nodes using consistent hashing, provide replication for durability, and expose a simple key-value API (get/put/delete) to clients. Discuss how you would scale the system horizontally, monitor performance, and tune it for both read-heavy and write-heavy workloads.