Anyone know why the predictive routing engine returns a 503 Service Unavailable when scaling concurrent sessions beyond 50 threads in JMeter 5.6? The load test targets the Genesys Cloud US1 environment. The script simulates inbound voice interactions using the standard REST API endpoints for queue management. The token is valid and permissions are correct. The error appears only after the thread count exceeds the limit. Lower thread counts work fine. The goal is to understand the call capacity planning limits for this specific configuration. The environment is Genesys Cloud US1. The JMeter version is 5.6. The error message is 503 Service Unavailable. The request path is /api/v2/interaction/interactions. The headers include valid OAuth tokens. The body contains standard interaction data. The issue is consistent across multiple runs. The latency spikes before the 503 error occurs. The websocket connections seem stable. The API rate limits are not being hit according to the logs. The problem is specific to predictive routing. Standard routing works fine at higher concurrency. The queue configuration is default. The agent count is 100. The skill assignment is correct. The issue needs resolution for performance testing. The documentation does not mention this specific limit. The error logs show gateway timeout occasionally. The JMeter response times increase significantly. The throughput drops to zero after the error. The test stops automatically. The error code is 503. The endpoint is /api/v2/interaction/interactions. The method is POST. The content type is application/json. The authorization header is valid. The issue is urgent for our current project. The team is blocked on this finding. The environment details are consistent. The JMeter config is standard. The thread group settings are simple. The ramp-up time is 10 seconds. The loop count is 1. The sampler timeout is 30000 ms. The result collector is default. The listener shows the 503 error clearly. The request body is minimal. The headers are standard. The issue is reproducible. The environment is US1. The JMeter version is 5.6. The error is 503. The endpoint is /api/v2/interaction/interactions. The method is POST. The content type is application/json. The authorization header is valid. The issue is urgent. The team is blocked. The environment details are consistent. The JMeter config is standard. The thread group settings are simple. The ramp-up time is 10 seconds. The loop count is 1. The sampler timeout is 30000 ms. The result collector is default. The listener shows the 503 error clearly. The request body is minimal. The headers are standard. The issue is reproducible.