Right now you can do that through the web interface, on the status page.
I’ve got a wget script to do the same thing, but the larger point is well taken — the idea of using some kind of “fake” scripts to execute scripts.
Almost everything is there to do that — the only thing that’s missing is to have multiple transcode scripts. I’m still working on getting the sql-backed version up to the same place the gdbm version was. Once I’m there, I can start adding the neat stuff that I couldn’t with the gdbm version.
Like multiple users. Audioscrobbler. Events. That sort of thing.