One could always modify StumpWM at runtime, via SLIME, which is far superior to geiser. Common Lisp is also superior to Scheme (as a language) and SBCL/CCL are miles better than Guile in particular, Guile being one of the worst Scheme implementations out there.
For libraries, CL has the same problems as Scheme so it's a wash. I hit this keenly when interacting with the OS (POSIX calls, mostly).
At least that's the case in the newer "open" CL world; older CL code, especially commercial code, more often did target a specific implementation like Franz, which is more like the Scheme approach.
You doing work on Linux and need a very fast compiler? Use SBCL, 3rd party libraries continue to work (unless they're platform specific)
Doing work on OSX or Windows? Use CCL. 3rd party libraries continue to work (unless they're platform specific).
This flexibility is absent from the Scheme world, with the myriad slightly incompatible implementations and extremely fragmented library space. The emergence of Quicklisp for Common Lisp was so important that it alone KILLED Scheme for any sort of practical use. For most cases, one has not the slightest incentive to stray from using Common Lisp these days, and if he does, he better have a good reason.
What is the equivalent for Scheme? The SRFI hell? _shudder_
Your comment would be much better without this bit, which is entirely subjective.
Guile is a GNU project, which means that if you care about (say) Windows (gasp), you're shit out of luck.
The same used to be true about OSX 2-3 years ago, I haven't bothered to check again since then.
Coupled with the fact, that maybe 2-3? people are actively working on Guile (guess what platform they're focusing on), and that pretty much nobody is using Guile besides little toy projects, things are not looking good.
No, a really good Scheme implementation would be, Racket or Chicken. Hell, even Gambit is objectively better and more mature than Guile. I guess you're too blind to see that.
Are you using StumpWM on windows?