A low but stubborn rate of 502s at the load balancer and socket hang up errors between services turned out to be one race: Node closed idle keep-alive connections after 5 seconds while its callers kept reusing them.
Our live support chat dropped every connection that stayed quiet for a minute, with close code 1006 and no server error. Nginx's 60-second proxy_read_timeout and the load balancer idle timeout were closing idle sockets. A heartbeat plus proxy tuning fixed it.
Our blue-green release was supposed to make rollback instant. Instead Green's migration renamed a column while Blue still served all traffic, and Blue started returning 500s before cutover. Both colours share one database. We rebuilt the change as expand, migrate, contract across three releases.
Our flash sales pushed traffic to eight times normal in under a minute, and autoscaling arrived five minutes late. This guide covers what actually kept us up: the right balancing algorithm, pre-scaling around known events, queues for deferrable work and shedding load on purpose.