Queue for Hypervel

Documentation: https://hypervel.org/docs/queues

Differences From Laravel

• Laravel's DynamoDB failed-job provider is not provided. Use database or file storage for failed jobs.
• Custom job implementations must support JobLifecycleCallbacks to use CountCrashesAsExceptions, so cleanup completes before a concurrent retry starts. The base Jobs\Job already implements it; the Job contract is unchanged. See counting worker crashes.
• Timeouts unwind yielding job code with Swoole's CanceledException, starting its catch and finally blocks. Use finally for cleanup and rethrow cancellation if caught. The worker stops after timeout handling by default, with a bounded wait for cleanup; Worker::$killOnTimeout = false keeps other jobs running. See worker timeouts for cleanup and blocked-loop recovery.
• The protected enqueueUsing() callback receives the queue that owns the operation as its first argument. This lets deferred work borrow a fresh pooled queue connection after a database transaction commits instead of retaining a connection for the lifetime of the transaction.
• Positive per-message delays on SQS FIFO queues throw LogicException. Laravel silently omits the delay and sends the job immediately.
• RateLimited::releaseAfter(0) requests an immediate retry. Laravel treats zero as absent and uses the limiter's computed retry delay.
• RedisQueue::scanQueueKeys() is intentionally omitted. Queue discovery streams keys on a held connection instead of materializing a physical-key array; override allQueueNames() to customize discovery.
• Redis bulk dispatch uses enqueueBatch() and LuaScripts::bulk() for both standalone Redis and Cluster, storing immediate and delayed jobs together with an exact completion count. Laravel's RedisQueue::bulkOnClusterConnection() and LuaScripts::bulkPush() helpers are intentionally omitted.

Ported from: https://github.com/laravel/framework/tree/13.x/src/Illuminate/Queue