Add tests for connect-disconnect and error handling in SmartProxy

readme.hints.md

@@ -523,4 +523,63 @@ if (!record) {
 - **Consideration**: More frequent cleanup operations, but prevents queue backlog
 
 ### Migration Notes
-No configuration changes needed. The improvements are automatic and backward compatible.
+No configuration changes needed. The improvements are automatic and backward compatible.
+
+## Early Client Disconnect Handling (v19.5.13+)
+
+### Issue
+Connections were accumulating when clients connected but disconnected before sending data or during routing. This occurred in two scenarios:
+1. **TLS Path**: Clients connecting and disconnecting before sending initial TLS handshake data
+2. **Non-TLS Immediate Routing**: Clients disconnecting while backend connection was being established
+
+### Root Cause
+1. **Missing Cleanup Handlers**: During initial data wait and immediate routing, no close/end handlers were attached to catch early disconnections
+2. **Race Condition**: Backend connection attempts continued even after client disconnected, causing unhandled errors
+3. **Timing Window**: Between accepting connection and establishing full bidirectional flow, disconnections weren't properly handled
+
+### Solution
+1. **TLS Path Fix**: Added close/end handlers during initial data wait (lines 224-253 in route-connection-handler.ts)
+2. **Immediate Routing Fix**: Used `setupSocketHandlers` for proper handler attachment (lines 180-205)
+3. **Backend Error Handling**: Check if connection already closed before handling backend errors (line 1144)
+
+### Changes Made
+```typescript
+// 1. TLS path - handle disconnect before initial data
+socket.once('close', () => {
+  if (!initialDataReceived) {
+    this.connectionManager.cleanupConnection(record, 'closed_before_data');
+  }
+});
+
+// 2. Immediate routing path - proper handler setup
+setupSocketHandlers(socket, (reason) => {
+  if (!record.outgoing || record.outgoing.readyState !== 'open') {
+    if (record.outgoing && !record.outgoing.destroyed) {
+      record.outgoing.destroy(); // Abort pending backend connection
+    }
+    this.connectionManager.cleanupConnection(record, reason);
+  }
+}, undefined, 'immediate-route-client');
+
+// 3. Backend connection error handling
+onError: (error) => {
+  if (record.connectionClosed) {
+    logger.log('debug', 'Backend connection failed but client already disconnected');
+    return; // Client already gone, nothing to clean up
+  }
+  // ... normal error handling
+}
+```
+
+### Test Coverage
+- `test/test.connect-disconnect-cleanup.node.ts` - Comprehensive test for early disconnect scenarios
+- Tests verify connection count stays at 0 even with rapid connect/disconnect patterns
+- Covers immediate disconnect, delayed disconnect, and mixed patterns
+
+### Performance Impact
+- **Positive**: No more connection accumulation from early disconnects
+- **Positive**: Immediate cleanup reduces memory usage
+- **Positive**: Prevents resource exhaustion from rapid reconnection attempts
+
+### Migration Notes
+No configuration changes needed. The fix is automatic and backward compatible.