Hi Ervin! First of all, we are very grateful for such an extended comment. It's great that our article encouraged you to discuss it.
Secondly, we want to share our thoughts on this. As you may have noticed, we are aware of the minuses that exist in our solution. You've exposed them more broadly and suggested that if we had bigger projects, our solution clearly wouldn't be the best. That's right.
But so far it suits us for several reasons:
1)our customers (e.g., a chain of stores, of which there are about 170) are satisfied with it. Even the biggest disadvantages (service drops for a few seconds) does not motivate them to switch to a more expensive solution. Expensive in terms of moving costs (developer time).
2)moreover we didn't want to redo the whole deployment process unnecessarily and "retrain" the whole team.
Nevertheless, for any enterprise projects we will 100% consider such popular solutions as you suggested.
I seems like you did not need Kubernetes in the first place, you just went with it because everyone does.
I don't know for whom you are working for as an outsourcing company and what are the requirements from your clients, but those cons are just rough:
Scaling: I understand that not everybody is required to scale, your monolith can take a lot of beating, but usually this feature is a nice to have. We can let this slide for now.
Resource Limiting: now this can bite you in case you encounter some unexpectedly large amount of traffic. Again, I don't know for who you working for, as a client I would expect that my service would run fine even if the neighbors are chatty. It is kind of a no go in many cases.
No updates without downtime: Now this is problematic. This is unacceptable for most of the clients.
In conclusion I'm asking, why did you not consider a managed environment where you can simply just host your containers. There are services such as AWS AppRunner, Google CloudRun, Heroku Dynos or many others. They are just easy to use. Nobody would mind if you have your docker compose thingy for development and testing, but going into production with this, you just reinventing the wheel.
Nevertheless, if it works for you, than good luck with it :)