These all strike me as iffy use cases. What is more compelling is creating a language for some more-specific need. These are generally referred-to as Domain Specific Languages, or DSLs (not to be confused with trying to push high-speed internet over a twisted pair...)
I designed one and implemented a compiler and interpreter for it in the early 1980's. It's not all that hard. I had had one compiler construction course in college. I used classic tools Yacc/Lex/Prep and wrote it in C.
The language is (was? haven't followed) called VSL, or Variation Simulation Language.
The problem was this: in the early 80's auto companies were experimenting with variation simulation. It's simulating the build of complex mechanical assemblies so that the effects of dimensional variations can be analyzed. The technique was developed at Willow Run Labs during WWII, as part of the solution to the awful-quality airplanes they were building for the war. They gathered experts to fix the problem, and they used this technique. At the time, it was done by a room full of woman working Friden mechanical calculators...
So, in the early 80's there was some Fortran code written by a university professor that ran on a mainframe. I worked for a company that set out to commercialize it. My first task was to port it from the mainframe to IBM PC.
Two problems: Models were written in Fortran, and then linked against a library. Fortran is painful, for anything. It's especially painful for manipulating representations of 3D objects. And compiling and linking Fortran on a PC was slow! Half-hour builds! And that's just to find you had a syntax error and then rinse and repeat.
My boss wanted to build a "menu system" that engineers could design in. Keep in mind, we are talking 80's and this was just to be a scrolling text menu. Yes, there were graphics workstations, but this was a new untested product, and nobody was going to pop the $20,000 that they did for, say, finite element workstations. they wanted it to work on a PC so that we could more easily convince the auto companies to try it - make it an easier decision to give it a go.
He wrote up the menu system, and presented it to us in the conference room. He rolled-out a roll of paper the length of the conference table, and then it hung over both ends! I convinced him that the time for this approach had not yet come.... Sure, point and click on graphics - but he couldn't afford either the time or money for that development. But not that silly long-ass text menu!
The alternative was VSL. It was specifically-tailored to the task, it had "objects" of a sort - and by this I mean "3D objects". You could just pass a fender around in a function call, for example.
It didn't compile to machine code, but generated bytecode. I wrote an interpreter in Fortran, and so eliminated the costly link step. The Fortran program just read the bytecode into an array and interpreted it. Was it slow? No, it was fast as heck! That's because almost all the work was done in well-optimized library functions written in Fortran or even assembly in some cases. (I also talked my boss into hiring an actual mathematician who fixed our broken edge cases, and knew the right heuristics to speed things up.)
This made it much easier for engineers to create and use models. Now they wrote them in VSL, much more expressive to the task than Fortran. And in a minute they either knew they had a syntax error or were testing their model.
In a couple years, we went from a couple of pilot projects to like 50. Every auto company took it up. Boeing used to help re-engineer the FA-18. Today probably every car, airplane, and hard drive was analyzed using VSL. (Siemens wound-up with the product eventually, after a few acquisitions.) I don't know if VSA is still under the hood, or if it really has any practical use today: the models are now written using point/click/drag/popup stuff on drawings. What my boss new we had to eventually get to, but couldn't at the time.
Of the languages mentioned in the article, I suppose Go makes the most sense, because they had a specific need that wasn't well-covered by other languages. (I wonder what was thought wrong with Erlang, though?) But they are all languages with pretty broad use cases, and so it is less compelling than something more focused.