My configuration keeps failing correctly when we simulate high concurrent inbound traffic.
Background
We are load testing our Predictive Routing (PR) setup in the US-East region. The goal is to validate agent capacity calculations. We use JMeter to generate 500 concurrent calls targeting a specific Workforce Management (WFM) group. The Architect flow is simple: Inbound → Set Queue → End. No complex logic. API version used is v2.
Issue
When the call volume exceeds 200 concurrent sessions, the queue depth in the Admin UI shows zero available agents, even though we have 100 agents logged in and set to Available. The calls start queuing indefinitely. The GET /api/v2/routing/queues/{queueId} endpoint returns waitTime increasing rapidly, but agentCount remains 0. This suggests the system is not recognizing the available agents under load.
Troubleshooting
- Verified agents are in the correct group.
- Checked
GET /api/v2/routing/users/{userId}; status isavailable. - Reduced JMeter threads to 50; issue disappears. Queue depth works as expected.
- No 429 or 503 errors in the API logs during the test.
Is there a hidden rate limit on the PR engine that prevents agent assignment under high concurrent request load? Or is this a known capacity threshold for the US-East region?