This distinction is a more useful one that the article made. I love dockerfiles and immutability, but there are good cases for mutable containers, too.
asciidoc is my favourite too, thanks to Asciidoctor. I agree, asciidoc hits the sweet spots as a format. Been frustrated by the tooling lately though. I can see the huge effort put into Asciidoctor, and am thankful for it, but there are still big downsides i.e. no semantic html 5 output, difficult (or at least more difficult than necessary) integration with image generators, heavyweight (only ruby dependency on my entire machine). I imagine this just needs more time and resources put to it, as all these issues (except the ruby one) are open on GitHub.
I have no expertise, but this is my sleep. It turns out I have sleep apnea. I can "sleep" for 8 hours, but it does very little good. With apnea, you wake up as much as a few times an hour, getting very little REM. It took me years to consider apnea because I'm thin and somewhat fit. But all sorts can have it.
I have to disagree with other responses. I think you could make a career of this for the same reason management consultants can.
You could swoop in and "fix" some stuff and then leave the in-house team not understanding what you've done. That sounds profitable. You might even get called back to fix things a second time.
As with management consulting, I think it would at the system level tend to do more harm than good, even if you do good work and get paid well for it. I agree strongly with feoren that the code needs to reflect the in-house developers' mental models of the domain or everything will fall apart. If you fix things and then give them a bunch of processes and coding standards to follow, they will not do well and you will be thought of as some clueless architecture astronaut by them. But profitably.
I have been getting tired of node dependencies and not knowing what I'm running or when it will break. I make simple things for small number of users and started playing with just writing my own CGI in C++ or bash behind thttpd. This appears to be working very well so far and runs on a vanilla Linux install with deps just on system libs and gcc. With all the pain and unsafety of deps, this might actually make most sense. What new vulnerabilities am I inviting? Why aren't we still doing things this way? It seems... much easier, perhaps not to make it but to secure and sustain it.