Server-Sent Events in production: lessons learned after 2 years

we've been using server-sent events (sse) in production for our real-time notification system for about two years now, and i wanted to share some lessons learned. initially, we chose sse for its simplicity compared to websockets, especially for server-to-client only communication, and its native browser reconnection capabilities. we definitely ran into connection limits on our backend servers, which required careful scaling and load balancing, ensuring sticky sessions were in place. managing memory per connection also became a concern as we scaled past 50k concurrent users. also, while browser reconnection is great, tuning the backoff strategy to avoid thundering herds during outages was critical. overall, it's been a solid choice, but don't underestimate the operational aspects of managing many long-lived http connections, even with sse's relative simplicity.