Task lifecycle

Tasks move through four statuses:

• new: available for leasing
• pending: leased by a worker
• successful: completed successfully
• failed: permanently failed or expired

Typical flow

1. append() inserts a task in new.
2. next() leases a task, updates it to pending, and sets assignedTo.
3. The worker calls on_success() or on_failure().
4. on_success() updates status to successful.
5. on_failure() increments retries and either requeues or marks failed.

Leasing and visibility

If visibility_timeout > 0, a lease expiration timestamp is stored as leaseExpiresAt. Calling refresh() requeues expired leases by setting status=new and clearing assignment fields.

TTL expiry

If ttl is set, tasks that remain pending longer than ttl are marked failed during refresh().

Discarding

Tasks with retry count at or above max_retries are discarded during refresh(). The action depends on discard_strategy: - keep: keep tasks in place and mark them failed - remove: delete tasks, optionally copying to the dead-letter collection