There are a number of very good “niche” programming languages available. But “niches” are not all alike.
Scala has a “niche” in parallel software. It’s a good language and you can use it for other things, but it buys you the most in that area.
Haskell has a “niche” in scientific programming. It’s also a good language and can be used for other things, but it buys you the most in that area.
LISP has a “niche” in recursive programming, which is useful for certain types of mathematical problems and for AI, it can be used for other things, but framing common business problems in LISP can be more of a challenge than it’s worth unless you’re really good with it.
Smalltalk, on the other hand, is also a “niche” language, but its niches seem to have little in common at first glance. It’s used in prototyping, due to the speed of revising iterative software, and simultaneously in production automation, including high performance manufacturing such as semiconductor fabs. It’s used in DoD defense automation systems, as well as things like telecom, where most of Iceland’s cell phone system is run on a Smalltalk back end. Another, seemingly unrelated niche is epidemiology software, such as Kendrick, used for the Ebola outbreak.
The commonality to these niches that stands out has less to do with specific language aspects that make certain types of calculations easier to frame, but rather that in each case “shit has to work”, and it has to be written, tested and debugged productively.
It seems odd that CIOs, technical leads and developers would see a language that productive and that reliable as merely a “niche”.