In my experience, that kind of design would be bad. I prefer light themes, but whenever I've used some app in dark mode and switch to a light one, the experience is inevitably jarring, and completely faithful to all the sun-burns-my-eyes jokes made about light themes. But it's fine the next day.
One of the more important parts of the studies referenced in the NNGroup article the blog post in the OP is referring to is that they did experiments between people rather than subjecting people to different conditions consecutively. When the initial light-dark contrast isn't present, people's fatigue ratings were about equal for both modes.
Moreover, people perceived both modes equally hard to read, but were actually more efficient in reading lightmode text.
That advantage is driven largely by the sheer amount of light - some researchers did an experiment where they cranked up the brightness on a dark theme so the experienced brightness was the same and it was just as readable as a light theme. But then you're using a crapton of electricity since you're turning dark-hued pixels to be super bright.