LayerWalker

v2.74 - Performance Optimization

• Reduced CPU usage in network message processing by ~50-70% for active networks (10-100+ players)
• Eliminated redundant UI recalculations using a generation-based dirty flag cache (BuildNetLayerLines now only rebuilds when data actually changes)
• Removed per-message overhead: PruneSwitchRequestTables moved from message handler to periodic 30-60s cycle
• Cached player name lookup (UnitName + StripRealm) to avoid repeated API calls on every incoming message
• Reduced dedup table (commSeen) TTL from 120s to 15s, shrinking table size by ~8x
• Optimized OnUpdate fallback to use elapsed accumulation instead of per-frame GetTime() calls (~3600 → ~4 calls/min)
• Extended CountRecentNetMembers cache TTL from 2s to 5s to reduce redundant table iterations
• Consolidated 4 separate prune loops in RefreshNetRosterSnapshot into a single pass with direct key removal