The point of data scientists and the related roles listed in the article are not to just churn out the fun stuff, but to wade through the institutional and technical muck and mire it takes to bring the fun stuff to bear on a relevant business problem and to communicate the results in a way that people of all walks can understand.
But even in this day and age with ML being the new hotness, you will find people who are quite happy to work on infrastructure and don't have a huge amount of interest in training models themselves, and it is probably a lot easier to hire them than people who can do both, and you may get better results from actual specialists.
I suspect, if there are lots of relatively simple ML problems, then a generalist with integration chops will be more effective in getting them out quickly and "good enough". The specialist may take too long on models that are too heavy and impractical.
If there's one big ML problem (Google search, Netflix recommender, Amazon search, etc), where 1% additional makes a difference, then yes, specialist DS/modeler is probably preferred.
Larger, older org/heavier existing infra/more specialized culture will also tilt the scale towards specialists.
I also think it's unfair to specialists to say they will always overcomplicate things more than others, I've seen plenty of generalists with researcher envy do the same thing.