Discord error 4000: Unknown error

The gateway closed your WebSocket without a specific reason. Discord's own description is deliberately vague, and in practice 4000 is the code you see when the connection dies for transport-level reasons: a Discord gateway node was restarted or rebalanced, your host's network dropped packets long enough to break the socket, a proxy or load balancer between you and Discord timed out an idle connection, or the server simply hit an internal error while handling your session.

It is marked reconnectable, and your library will almost always attempt to Resume automatically using the stored session_id, resume_gateway_url and last sequence number. If the resume succeeds you get a RESUMED event and replayed missed events; if it fails you get an INVALID_SESSION and a fresh IDENTIFY. A handful of 4000s per day on a long-running bot is normal. A storm of them points at your network or at an intermediary.

// WebSocket close frame (no JSON body)
code: 4000
reason: "Unknown error"

// what your library logs (discord.js style)
[WS => Shard 0] Connection closed with code 4000, attempting resume...

Keep the event loop responsive so heartbeats are always on time, run the bot on a host with stable egress, and enable your library's debug logging for shard lifecycle so you can tell resume from re-identify. Track disconnect counts as a metric rather than ignoring them.

Distinguish disconnect rate from disconnect existence: some 4000s are the cost of a long-lived socket, and a resume makes them invisible to users. The gateway timing rules that govern resumes and heartbeats are on the Discord limits page, and a bot that dies rather than resumes after 4000 should be checked against the offline checklist in Discord bot not responding.