I suggest skimming through a few pages of Planet Clojure and Planet Lisp, to get a sense of the apps people are writing. As a rough rule of thumb, if one excites you much more, that could be the better choice.
If you end up going with CL, you'll likely have a better time asking newbie questions on LispForum (or mailing lists like the LispWorks one), rather than on Usenet. And there are alternatives to SBCL and CMUCL, such as ECL and LispWorks: http://common-lisp.net/~dlw/LispSurvey.html
As for prototyping-then-rewriting, I think there are a few situations:
1) nontechnical reasons (such as, the maintainers can't be expected to maintain a Lisp app)
2) because the Lisp implementation is missing some convenient functionality, and it isn't easy to call an existing one in another language
3) the Lisp doesn't interoperate well with a given technical ecosystem
I think the probability of rewriting is less with Clojure, due to its Java interop, long-term goals of interop with other platforms (like .NET), and the increasing number of teams using it.