Queue Retries

Keep these execution choices in mind when using the linked guides:

• Use the built-in queue for sequential background work. Retries block later queue items; use scheduling for long recovery waits.
• Retry delays keep the Durable Object active. Choose retry budgets with execution cost and latency in mind.
• Queued items are removed after their retry budget is exhausted; there is no built-in dead-letter queue. Record failures explicitly when the application needs recovery or auditing.
• The selective retry predicate is available on this.retry(), not serialized queue or schedule options. Handle non-retryable errors in those callbacks.

See state-scheduling.md for choosing schedule modes and persisting application state.