Hot keys occur when one cache key receives disproportionate traffic. A viral tweet or trending product can overwhelm a single cache shard.
Solutions:
Key splitting: Append random suffix (key-, key-, ..., key-N). Read from random suffix.
Read replicas: Replicate hot keys to multiple nodes
Local caching: L cache in application memory for hottest keys
Detection: Monitor key access patterns. Alert when single key exceeds threshold (e.g., QPS).