When systems get backed up and hit concurrency limits, background tasks can sit in the queue for too long. If a job is highly time-sensitive—like sending a verification email or processing a real-time webhook—executing it late might be worse than dropping it entirely.
Time-to-live (TTL) limits can now be configured for task runs. If a run isn't dequeued before its specified timer runs out, it automatically expires and is never executed. The configuration allows defaults to be set at the global project level, overridden on a per-task basis, and overridden again for individual triggers. This prevents stale runs from piling up or executing when their output is no longer relevant.
The documentation for these capabilities is being added across the core guides and configuration references, landing alongside the package updates in PR #3196.