feat(apps/hermes): add connection timeout for SSE & WebSocket connections

[cctdaniel]

Summary

• Added a 24-hour connection timeout for SSE endpoint
• Added a 24-hour connection timeout for WS endpoint
• Each client connection will automatically disconnect after 24 hours of continuous connection, with a proper error message sent before closing.

Rationale

• Prevents resource leaks from indefinite connections
• Ensures proper cleanup of long-running connections
• Each client gets their own independent 24h window from their connection time
• Provides clear error messaging to clients about connection termination

How has this been tested?

• Current tests cover my changes
• Added new tests
• Manually tested the code

Testing steps:

1. Verified that each new SSE/WS client connection gets its own independent 24h timeout
2. Confirmed that the timeout message is properly sent before connection close
3. Tested that the stream properly terminates after the timeout period
4. Verified that existing functionality (price updates, error handling) continues to work during the connection lifetime
5. Tested that multiple concurrent clients have independent timeouts based on their individual connection times

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

6 Skipped Deployments

Name	Status	Preview	Comments	Updated (UTC)
api-reference	⬜️ Ignored (Inspect)	Visit Preview		Apr 24, 2025 11:46am
component-library	⬜️ Ignored (Inspect)	Visit Preview		Apr 24, 2025 11:46am
entropy-debugger	⬜️ Ignored (Inspect)	Visit Preview		Apr 24, 2025 11:46am
insights	⬜️ Ignored (Inspect)	Visit Preview		Apr 24, 2025 11:46am
proposals	⬜️ Ignored (Inspect)	Visit Preview		Apr 24, 2025 11:46am
staking	⬜️ Ignored (Inspect)	Visit Preview		Apr 24, 2025 11:46am

[ali-behjati]

Please address my comment before merging and bump the version. Also please update the docs (for SSE at least) to document this behaviour.

[tejasbadadare]

nice! one thing i'm wondering about is that there are many clients out there that maintain open connections to hermes. when we deploy this change, they'll all reconnect and start their connection TTLs at the same time. then every 24h there will be a thundering herd of reconnections. we could add a jitter to the 24h to avoid this (give them a variable extra 1-5m after the 24h).

in reality, many of them will drop their connections naturally due to network conditions before 24h elapses, and hermes can probably handle the thundering herd anyway, but just something to be aware of.