But that's not how languages have typically been selected for 15 years or so. The direct vocational skills argument has held much more sway.
Basic use of scheme requires you to think recursively. If you have that, scheme gives enormous power for a small investment of learning, thinking and typing. If you don't have that, you need to deal with that before you can use scheme.
You could take a group of elite students with a math/physics background and have some confidence that their grounding in mathematics and physics will have prepared them, or that it will be a short and healthy jump for them.
But it would be bigger hurdle for other groups. Python is a useful because you can ease people into this kind of thinking. And people who aren't there yet can still get things done.
Regarding SICP, even in early chapters, it requires you to understand maths concepts that wouldn't be available to you even if you were a professionally-employed person returning to to university. The Little Schemer path would be a better intro for groups without that.
But every place isn't MIT, nor is every course a common core requirement for a college or department with diverse majors. Does a civil engineering student really need to implement lazy evaluation?
Felleisen's paper and subsequent book as well as the work of the PLT group show the current state of Scheme as a pedagogical language.
http://www.cs.brown.edu/~sk/Publications/Papers/Published/ff...
Nope, MIT has an introductory computing course that logically goes before 6.001-4 or the new 6.01-2. For a long while they didn't have one due to resource constraints, they dropped this service course about the time 6.001-4 was developed, but other departments had their own more domain specific introductory courses (see below for more details).
6.001 was canned and Scheme completely purged from the undergraduate curriculum for political reasons, after the Department, and likely School of Engineering and the Institute itself panicked when the dot.com bust more than halved enrollment, after it had been 40% of the undergraduate class for decades. Oh, yeah, after the Institute had also blown a quarter billion dollars in a not very functional new building for the EECS research labs.
"... and were there as because of a common core curriculum."
That ignores that MIT's EECS department has for a very long time insisted that EE majors have serious programming experience, and CS majors the same for hardware. 6.001 programming, 6.002-3 hardware, 6.004 both, all majors had to take all 4. This was true but less coherent in the '70s before that common core was developed, and it's still true with 6.01-2 which combine both. MIT's EECS department started as an EE department, this is not an unusual approach for such, vs. ones that started from the math side or otherwise independent of EE. If you want pure CS, you shouldn't go to MIT or U.C. Berkeley, but to CMU or Standford (or so I gather about the latter three).
"Does it really make sense to teach such people to write language interpreters in an introductory course?"
Yes, if you want them to really understand what it's all about. Since it's Lisp, that's also much less intense than it normally is, essentially all the colossal parsing messes with other language families can be skipped thanks to its S-expression syntax, and you can focus on the very basics.
As for e.g. Civil Engineers, MIT's Civil and Environmental Engineering department has its own 1.000 and 1.001 courses for its type of computing, and as I remember that was true way back when 6.001-4 was developed. In the late '70s 4 or so engineering departments got together and got VAX11/780, which they used for their own education etc. purposes, while EECS later procured first a DECSYSTEM-2060 (which had a CPU twice as fast, but a max of 1 MiB address space), then/in addition HP 68000 workstations (eventually Project Athena took over this sort of duty, I think).
Finally, whatever the pedagogical merits of Scheme and/or SICP/6.001, Python is now winning due to fashion rather than merit (although it does have some merits compared to other languages that could and are often used like Java).