My guess is still that the best way to present the implementations on
the Scheme.org front page is to divide them into 2 or 3 categories.

We should have a separate, more detailed implementation browser where
one can sort and filter by various criteria, peruse benchmarks and
conformance matrices, etc. But the first impression counts, and should
be something relatively easy to digest.

The most common use cases for Scheme are probably:

- The fastest, most full-featured implementations with threads and
debugger: Gambit, Chez Scheme, MIT Scheme.

- Something light and easy to embed for glue code and scripting:
Chibi-Scheme, S7, TinyScheme.

- Something in between, for substantial portable programs that don't
need especially high performance: Guile, Gauche, Chicken.

However, as far as I know Guile, Gauche, and Chicken all have full
continuations and fully general tail calls, so in that respect they are
no different from the first group.

I don't know about the situation with the tiny interpreters.
https://github.com/lambdaconservatory/femtolisp/blob/master/tiny/lisp.c
from Jeff Bezanson is a pedagogical tree-walking Lisp interpreter in
only 1000 lines of C that nevertheless has lexical scope and proper tail
calls. This interpreter does not have continuations, however.