Quick question about predictive routing capacity limits.
Context:
Running a load test from Singapore using JMeter 5.6.2. Trying to simulate 300 concurrent inbound calls hitting a predictive outbound campaign. The flow is simple: connect to agent, transfer, hangup. No complex logic.
Question:
When the thread count hits around 250, the /api/v2/floors endpoint starts returning 503 Service Unavailable. The WebRTC tokens are generating fine (200 OK), and the WebSocket connections stay open. It seems like the routing engine itself is rejecting the load before the call even tries to match an agent.
I checked the org limits and we are well under the licensed capacity. The error happens consistently at the same throughput. Is there a hidden rate limit on predictive campaign initiation that differs from standard inbound routing? Or is this a known issue with the JMeter plugin not handling the async callback correctly?
Any insights would be great. The test script is pretty standard OAuth2 bearer token flow.