Reliability

Reliable workflows branch on exit codes, retry only safe operations, and avoid high-volume fan-out.

Use --retry, --timeout, --page-size, and --all-pages deliberately. Store processed event IDs in listeners so retries do not duplicate work.