I'm trying to move slow work off my request path and need a recommendation for durable background job processing

If you want durable background job processing—especially to move slow work off the request path—the safest recommendation is usually:

Best general choice: a real job queue with durable storage

Use a queue-backed worker system rather than in-process threads/tasks.

Strong options by stack

• Python
  • Celery with Redis or RabbitMQ
  • RQ for simpler needs
  • Dramatiq as a modern alternative
• Node.js
  • BullMQ with Redis
  • Agenda if you want MongoDB-backed jobs
• Ruby
  • Sidekiq with Redis
• Java / JVM
  • Spring + RabbitMQ/Kafka
  • Quartz for scheduled jobs, not ideal as a generic durable queue
• .NET
  • Hangfire for persistent background jobs
• Go
  • Use a queue like NATS JetStream, RabbitMQ, SQS, or Kafka, with worker processes

What “durable” should mean

For request-path offloading, you want:

• Jobs persisted before acknowledging the request
• Retries with backoff
• Idempotent processing
• Dead-lettering / failure handling
• Separate worker processes
• Observability: retries, lag, failure rates, job history

My practical recommendation

If you want something reliable and common:

• Celery + RabbitMQ if you’re in Python and need stronger queue semantics
• Sidekiq if you’re on Ruby
• Hangfire if you’re on .NET
• BullMQ if you’re on Node.js and Redis is already acceptable
• SQS + worker service if you want managed durability and less ops

If you’re trying to keep it simple

A good minimal pattern is:

1. Web request validates input
2. Persist a “job record” in your DB
3. Enqueue job or let workers poll the DB/outbox
4. Return 202 Accepted with job ID
5. Worker processes job asynchronously
6. Client checks status via API or webhook

This is especially good if you need:

• transactional consistency
• auditability
• recovery if the queue is temporarily unavailable

Avoid

• In-memory queues inside the web process
• Fire-and-forget threads/tasks from request handlers
• Anything that disappears on process restart unless that’s acceptable

My quick default pick

If you didn’t mention a language, I’d recommend:

• SQS + worker service for managed durability and simplicity, or
• RabbitMQ + workers if you want more control and portability

If you tell me your language/framework and whether you already run Redis, RabbitMQ, Kafka, or a cloud platform, I can give you a much more specific recommendation.