I would be able to speak to the allegedly “impractical” nature of Scheme if only I knew why Cisco hired Kent Dybvig, the principal developer of Chez Scheme.
I would also like to know why Beckman Coulter Life Sciences supported the development of Swish, an extension of Chez Scheme that provides Erlang-inspired message passing.
The impression of being 'impractical' came from the Scheme reports, which for a long time, with the exception of the controversial R6RS, only standardized a relatively simple/limited language (example: no error handling).
why Beckman Coulter Life Sciences supported the development of Swish, an extension of Chez Scheme
Having worked at Beckman¹ from 1978-2022, I strongly suspect that the support of a Scheme-based system was due to the educational background of several of the senior developers in that group of the Life Sciences software development team.
¹ Beckman Instruments
-> SmithKline Beckman
-> Beckman Instruments
-> Beckman Coulter
-> Beckman Coulter acquired by Danaher 2011
-> Beckman Coulter Capillary Electrophoresis business moved (2013) to AB Sciex (also a Danaher company)
Probably impracticality via ecosystem. Chez scheme is fast, but more limited library wise than common lisp and much more limited than Python in that manner. Note that I like Chez scheme and think it's really cool.
Gambit and Scheme have been used for health applications. Here's one paper on that subject, from 2013: https://ecem.ece.ubc.ca/~cpetersen/lambdanative_icfp13.pdf
I would mention the use of GNU Guile to build Guix, which has had considerable uptake. Guile has been used to build other Linux programs.
Admittedly, Scheme isn't widely used. But impractical? No!