Try deploying in a different region (eg, US East).

I tried that. It did not resolve it.

I already moved to US East (Virginia) and the issue persists. The specific symptoms are:

• /socket.io/ HTTP requests timeout at exactly 45 seconds returning 0 bytes in HTTP logs
• Network Flow Logs show repeated TCP_OVERWIN errors
• Socket disconnects with transport close on ALL clients including mobile on different networks
• server.keepAliveTimeout = 60000 and pingInterval: 10000 are set but issue persists

Can you tell me the exact idle timeout value configured on your proxy/load balancer for WebSocket connections? I need to set my pingInterval below that threshold.

### 1. Disable WebSocket Compression (Server-Side)

To prevent proxy-level windowing errors (such as TCP_OVERWIN), you need to explicitly disable the perMessageDeflate option in your Node.js server configuration.

```javascript

// Server-side (index.js / server.js)

const io = new Server(httpServer, {

cors: {

origin: "*",

},

// Critical fix for TCP_OVERWIN issues on Railway

perMessageDeflate: false

});

```

### 2. Force WebSocket Transport (Client-Side)

You can bypass the HTTP long-polling phase—which is prone to 0-byte hangs when the proxy desynchronizes—by forcing a direct WebSocket connection from the client.

```javascript

// Client-side

const socket = io("https://your-service.up.railway.app", {

transports: ["websocket"], // Skip polling entirely

upgrade: false

});

```

### 3. Verify Environment Stability

If the connection issues persist, review your environment configuration in the Railway Dashboard:

* TCP Proxy Configuration: Navigate to Railway Dashboard > Settings > Variables and ensure you haven't manually enabled a TCP Proxy if your application only requires standard HTTP.

* Sticky Sessions: If you are running multiple replicas, you must enable Sticky Sessions in your Railway service settings to maintain Socket.io state consistency across instances.