Not sure where you're getting those requirements from — message queues were called out specifically in TFA and async workflows are trivial to build on top witho

Not sure where you're getting those requirements from — message queues were called out specifically in TFA and async workflows are trivial to build on top without instantly reaching for Kafka The "just use postgres" mantra can take you shockingly far. If you're at a scale where postgres doesn't actually make sense, then you have enough money to figure out and migrate to the best tech for your use case. Anecdotally, we're processing a few hundred million messages a day through postgres and never have a hiccup. All of our developers already know how to use it so the learning curve and ops burden is essentially zero. We'd have been foolish to pick something else