Jobs and retries

Music Delivery returns before storage verification, catalog coordination and media processing finish. Durable jobs make that work observable and recoverable without holding an HTTP request open.

Durable work

A batch contains item and job state with immutable public IDs. Workers claim leased jobs, publish results only for the current attempt and reject stale result context. Fetch current state with getMusicDeliveryBatch.

States

Queued, processing, ready, failed and other documented states describe the specific entity being observed. Do not infer that a ready upload means a ready listening artifact, or that one failed item invalidates every unrelated item.

Retryable

Retry only when the occurrence or response marks work retryable. Honor Retry-After, use jitter and preserve the same idempotency identity. Metadata correction, rights action and checksum mismatch require a state change before retrying.

Terminal

Exhausted or non-retryable work needs a human or operator decision. Preserve issue evidence and correlation IDs; do not create loops that continuously resubmit an unchanged invalid manifest.

Idempotent retry

Use retryMusicDeliveryBatch only for audited retryable failures. Replaying an accepted submit with the same payload returns the existing batch; changing content under the same sender/message key conflicts.