Predictive Routing queue latency spike at 500 concurrent calls

• Genesys Cloud US1 region
• Architect flow: Simple queue with predictive routing enabled
• Load test tool: JMeter 5.6.2
• Concurrent agents: 500
• API endpoint: /api/v2/routing/queues

Stuck on a weird latency issue with predictive routing during load testing. When JMeter ramps up to 500 concurrent agents, the time between an agent becoming available and receiving an offer jumps from 2 seconds to over 15 seconds. The API calls to update agent status return 200 OK, but the queue metrics show a high ‘Wait Time’ even though there are plenty of calls in the queue.

The predictive routing settings are default. No custom skills or languages configured. The issue only happens when the concurrency hits this specific threshold. Below 400 agents, everything works fine. I am checking the API rate limits but they seem okay. Is there a specific configuration for predictive routing that handles high concurrency better? Or is this a known limitation with the WebSocket connections for agent state updates?

You need to verify your Data Action timeout settings and ensure the webhook payload isn’t causing the routing queue to stall while waiting for external responses. Check if predictive routing is configured to bypass the Data Action during high-load scenarios to prevent queue latency spikes.