Philipp Sumi

Yeah, concurrency would make or break Pro for us due to race conditions. Unfortunately, I can't open the link, I assume the repo is private? > I see is that...

Unfortunately not - your first assumption is correct. I have only one queue (called `foo`). The strings `a` and `b` are indeed the group IDs.

A big difference I'm seeing is that you are creating your worker _after_ you've enqueued your jobs. In my case (which probably is more real-life), the worker is being created...

If that helps, here's my full queue service that works on top of BullMQ. Will hopefully transpile out of the box after fixing the two TODOs: ``` import { JobExecutionOptions...

I see your point, but I think your setup may be too simple to reveal the issues I'm, seeing. My expectations was that - I should be able to enqueue...

I adjusted your example a little to add some randomness into the delays and I think I got it to break. Some observations: - Enqueueing the 100 jobs takes about...

I'm using hosted redis for local development, maybe that one is throttled a little (using the free offering). That still wouldn't explain the different behavior though... I sent you my...