For strict filtering, where events are far and few between, the connection drops due to idle timeout on alb. We need to experiment ways to keep the connection alive.

As alb time out extension is not an option, since the filtering may cross the timeout anyway.

• Discuss with the team possible options here:

• We can explore a heartbeat message option. Maintain the time when the last SSE was sent. If it is close to an hour (say 55 mins), send a "dummy" heartbeat message as SSE.
• Other suggestion?

• Implement and deploy to dev after discussion

QA / Acceptance criteria
Open a connection to realtime dev after deploying. Apply a very strict filtering (or unsupported namespace, etc.) such that events are likely not coming for the next hour. See that the connection survives.