I've gone though all of these at different scales. What I find these days are: 1. databases have gotten very good where a separate queue infrastructure isn't wo

I've gone though all of these at different scales. What I find these days are: 1. databases have gotten very good where a separate queue infrastructure isn't worth it, 2. databases provide much better observability (you can query the contents). If you really do want a queue, you're still better off using a stream (e.g. Kafka) for multiple producers and/or consumers. Using a database table as a 'transactional outbox' is a poor-man's Kafka which works well enough without the additional infrastructure for most scales. Redis taking some of the duty as you mentioned and microservices/distributed systems being less fashionable likely also factor into it.