I like the simplicity!
I used to combine RabbitMQ and Celery for async tasks. Mostly because it's what I learned to do since it was already in use at my first job. But Celery is such a pain -- or, at least, it was at that job. So many configuration options, different places where they were stored in different versions[1]. Weird errors. Poor documentation (at least at the time?)... I just started going for something simpler: rq and rq-scheduler with redis. For most of my use cases it's more than enough.
I've got to say that your approach has gotten me thinking of maybe simplifying everything even more. We'll see where I end up. In about 4 weeks I'll have to introduce asynchronous tasks in our current project, and though I was thinking of going the rq-way, your article has given me food for thought.
Other than that, your backend stack is mostly like what we use for our projects. We also use plain old docker + docker-compose, with the small difference that we have a somewhat hacked-together system I built with bash several years ago to extend docker-compose's functionality a bit and make every component somewhat more reusable between projects and easier to fine-tune on a "per-environment" (development, staging, production, etc) basis. We also use nginx, but your article has convinced me to look into alternatives.
Once again, thank you for your articles, they're a joy to read and think about!
[1] To be fair, that job had several aging Django codebases and I know most of them are still stuck with Python 2 and outdated Celery and django-channels versions. I constantly kept pushing for us to get rid of technical debt, but we never got to it...it's part of the reason why me and a mate left it for our own endeavors together.