ai": 2] io.temporal.internal.worker.BasePoller : Failure in poller thread Workflow Poller taskQueue=“PROCESS_EMAIL_TASK_QUEUE”, namespace=“labs-ai”: 2
io.grpc.StatusRuntimeException: UNAVAILABLE: Subchannel shutdown invoked
This is usually transient and the SDK retries UNAVAILABLE grpc errors
What is the ratio long_request/long_request_failure?
Do you have a proxy or load balancer? if so can you check the logs, and temporal server logs as well.