One of the most critical metrics for live operations is concurrent users (CCU), but tracking it accurately in real-time presents unique challenges. We need to distinguish between truly active players versus those who are logged in but idle, handle server instances that may report at different intervals, and aggregate data across multiple regions without double-counting. Our current system updates CCU every 30 seconds, but during peak hours or special events, we need even more granular visibility. I'm particularly interested in how others handle edge cases - players who disconnect without proper logout, mobile players who background the app, or users whose sessions span server restarts. What metrics do you track alongside CCU to get a complete picture of player activity? How do you detect and respond to unusual spikes or drops in real-time user counts? Any recommendations for efficient data structures or aggregation patterns?