Poller Autoscaling dynamically adjusts the number of concurrent polls a Worker makes to pick up tasks, using feedback from the Temporal Service. This reduces manual poller tuning and can help tasks start sooner when polling concurrency is limiting throughput. It does not change Worker count, compute capacity, or task execution concurrency.
All Temporal Cloud namespaces will have Poller Autoscaling enabled by default by December 20th, 2026. Workers using default poller settings on a supported SDK version will auto-enroll with no configuration change required. Explicitly configured poller settings are preserved. Supported SDK versions: Go (>= 1.47), Java (>= 1.38), Rust (>= 0.6.0), with Python (1.36.0), TypeScript (1.26.0), .NET (1.22.0), and Ruby (1.11.0) auto-enrolling upon release.