Agreed. On the last product I worked on we had a custom UI with no translation/internationalization support and we ended up spending a total of 2 or 3 person-months adding the capability. But then there was also a huge amount of time we had to spend also giving the translators sufficient context to translate the UI because we kept getting inconsistent translations. And there were conflicting regulatory requirements for labeling and units in some places that we had to handle by displaying different assets.
We came to the conclusion that we could have spent a month implementing a tool for the translators and probably saved 2 or 3 months of back-and-forth and screenshotting.
While it's technically practical to architect for translation and internationalization, it's frequently out of scope for a first version of most products. If you're selling something for the US market, it's difficult to justify bumping your time-to-market to add support for something that you won't need for your first release. And that's not getting in to the additional complexity that it adds to the software system to have that support.
It's difficult for me to take someone seriously when they're making statements about internationalization not requiring any additional resources. It just tells me that they have no experience in any of the fields where it does require additional resources. And reading the article, I can see that the "experts" are all from web development companies where there's a framework that exists and you don't have to worry about justifying your use of Software of Unknown Pedigree to an auditor. Or where it doesn't really matter if your translation or internationalization framework doesn't work 100% of the time because it's unlikely to kill someone if a control on your UI is labeled incorrectly.