While I mostly agree, I don't think it's so black and white on the enforcement part and I actually think that a lot of recent developments actually put holes into this.
The typical example is that when you have to explain something is the job of X and Y. Usually this means that X and Y are breaking those boundaries. Just make a semi-private (or even public) API only used for that thing and you have a broken boundary again. Or push it on a message queue, etc.
I think, it certainly helps, but then again it doesn't prevent it. Having modules you have the same effect.
So in the end you get more spots where things can get wrong and more operational complexity. If you stick to using them right, I think you can also stick to using modules right with less complexities, better performance, easier debugability, fewer moving parts.
Hackers will find a way to do hacky things everywhere. ;)
Also this whole discussion reminds me of Linus discussing how he thinks micro kernels add complexity a very long time ago. Not sure if they should be considered modules or microservices though.
Sharing global state and so on while maybe it shouldn't be done lightly, without thinking about it can and does make sense. And in most modern environments it's not like the most quoted issues can happen too easily.
Also I strongly agree with pointing out that this is actually a niche topic. It's mostly big because that niche is where probably the majority of HN and "startup" people spend their time.