I just checked in a couple of changes., addressing the two biggest annoyances we had when using CC.rb @Scribd.
1. There is an interplay between builder cleaning old builds and dashboard listing builds that are being cleaned. Before these changes, it resulted in "phantom" unfinished builds for revisions that in fact were built days ago. This was caused by lazy initiation logic of the artefacts directory. Now, it's moved to constructor, is explicitly invoked when starting an actual new build, and life is good. Although you can still get an occasional HTTP500 from the dashboard (once in a couple of weeks) because of the same synchronization issue.
2. Even with 7 characters, Git revision numbers are not sufficiently unique. So, now CC.rb stores full revision guid in the file system, but displays only the first 7 characters thereof.
The next biggest annoyance for me is the API preformance. With 30 CCMenu's hitting it every second, lack of caching forced us to clear old builds often. Effectively, we only have build history for 3 days. Not a big deal, but meh. Same goes for the home page and RSS feed.