Reconnection

On by default. When the engine drops unexpectedly:

The namespace moves to Reconnecting(attempt).
The engine retries with exponential backoff plus jitter, from reconnectionDelayMs up to reconnectionDelayMaxMs.
On reopen, every namespace that was still open — you called open() and not close() — sends its CONNECT packet again automatically.
Your events(name) collectors stay subscribed throughout. There is nothing to re-register.
State returns to Connected.

The jitter is randomizationFactor and exists so a server restart does not bring every client back simultaneously.

The one thing you must do yourself

Re-emit your server-side subscriptions. The library reconnects the connection; it does not replay your outbound messages. If your server tracks rooms or channel membership per connection, that state is gone after a drop and only you know how to rebuild it.

socket.connectionState.collect { state ->
    if (state is ConnectionState.Connected) {
        socket.emit(Subscribe(buildJsonObject { put("pair", "BTC-INR") }))
    }
}

Because Connected is re-entered after every successful reconnect, this handles both the first connect and every subsequent one. Replaying emits automatically was considered and rejected — the library cannot know which messages are idempotent, and replaying a non-idempotent one is worse than dropping it.

Servers with connectionStateRecovery enabled (Socket.IO 4.6+) restore rooms themselves. Sockit does not implement the client half of that yet, so assume you need to re-subscribe.

What to do per situation

Situation	What to do
Transient network drop	Nothing. Re-subscribe if your server needs it
`Failed` — auth rejected, ping timeout, TLS	Terminal. Fix the cause, then `close()` and connect with new options
`reconnectionAttempts` exhausted	Retries stop. Call `open()` to start again
Intentional shutdown	`close()` sends a force-close. No reconnect is scheduled

Tuning

socketOptions {
    reconnection = true
    reconnectionAttempts = Int.MAX_VALUE
    reconnectionDelayMs = 1_000
    reconnectionDelayMaxMs = 5_000
    randomizationFactor = 0.5
}

Defaults retry forever, which suits a foreground app that should recover on its own. Set reconnectionAttempts to a finite number if you would rather surface a failure to the user, and reconnection = false to manage retries entirely yourself.

To stop retrying without disconnecting — backgrounding, say — use pauseReconnect() rather than reconfiguring.

Keyboard shortcuts

Sockit

Reconnection

The one thing you must do yourself

What to do per situation

Tuning