We use a home-grown solution, and try to keep it pretty minimal. For example, we can run all our pipeline steps serially and still get all the work done plenty

We use a home-grown solution, and try to keep it pretty minimal. For example, we can run all our pipeline steps serially and still get all the work done plenty fast, so, for now, we're keeping that parallelism can of worms firmly shut.