Network latency above 200ms causes worker desync and phantom retries

When network latency exceeds 200ms during broker heartbeats, workers desynchronize from the broker, causing phantom retries that result in duplicate message enqueues. Root cause is TCP keepalive mismatch between worker and broker.

Set broker.heartbeat=30s in broker configuration. Use sticky connections via ConnectionMiddleware to maintain consistent worker-broker connections and prevent desync.