We use worker threads directly to process large unorganized (for the browser) datasets and do some deductions before it hits the store.

We also have a worker thread blocked on a redis channel that acts as a queue.