Does anyone know the hard limit for concurrent WebSocket connections per Architect flow instance? We are running a load test using JMeter to simulate 200 concurrent calls hitting a standard IVR menu structure. The flow works fine for the first 50 calls, but after that, the media stream drops unexpectedly.
The Genesys Cloud UI shows the interactions as “In Progress,” but the client-side WebSocket connection terminates with a 1006 error code. No specific API error is returned from the /api/v2/architect/flows endpoint, just a silent disconnect on the media layer. This happens consistently at the 51st concurrent session.
We are using the latest JMeter plugin for Genesys Cloud and the standard WebSocket handshake sequence. The payload size is minimal, just basic DTMF digits being sent. We have checked the firewall logs, and nothing is being blocked on port 443. The issue persists even when throttling the DTMF send rate to once every 5 seconds.
Our environment is in the Asia Pacific region, and we are on the standard platform API version. We suspect a rate-limiting issue on the WebSocket layer rather than the REST API, but the documentation is vague on concurrent session caps per flow. Any insights on where to look next?