We're designing a public-facing API to expose real-time game metrics to our analytics dashboard, third-party tools, and partner services. The API needs to serve current CCU counts, live leaderboards, active event participation, server status, and various performance indicators - all with minimal latency. The challenge is balancing freshness of data with API performance and cost. We're considering WebSocket connections for true push-based real-time updates versus polling with short TTL caches. Authentication, rate limiting, and data privacy are also major concerns, especially when exposing player-related metrics. What architectural approaches have you used for real-time metrics APIs? Do you implement different tiers of data freshness for different use cases? How do you handle versioning and backward compatibility when metric definitions evolve? Any recommendations for monitoring API health and detecting abuse patterns? Would appreciate insights on both technical implementation and business considerations!