how to optimize real-time data fetching without overwhelming the backend?

hey everyone, i'm wrestling with a common real-time system challenge, efficiently getting fresh data to users without burning out our servers. we're seeing a lot of concurrent connections and frequent updates, and our current polling approach is starting to hit its limits. i've been thinking about WebSockets or Server-Sent Events, but i'm curious about what strategies others have found effective. are there specific techniques for managing connection counts, throttling updates, or batching data on the client-side that have worked well for you? looking for real-world experience here