Encountering 503 Service Unavailable errors when load testing Architect IVR flows with concurrent WebSocket connections exceeding 500 per second. Environment is Genesys Cloud US1 using JMeter 5.6 with the custom WebSocket sampler plugin. The flow uses a simple menu with three options and routes to a queue. When concurrent sessions reach 450, the latency spikes to over 2 seconds, and at 500 concurrent connections, approximately 15 percent of requests fail with HTTP 503. The error occurs specifically during the initial WebSocket handshake phase before the flow logic is fully engaged. We have verified that the underlying infrastructure can handle the load, so we suspect this is a platform-side throttling limit for new Architect session initiations. The JMeter script uses a ramp-up period of 60 seconds and holds the load for 300 seconds. No custom API rate limiting is configured on our tenant. Is there a known hard limit on concurrent Architect session creations per minute or per second that we are hitting? We need to validate if this is expected behavior for our capacity planning or if there is a configuration setting in the Architect flow that can optimize the connection handling. The logs show the 503 error coming directly from the Genesys Cloud edge servers before any flow logic is processed.