On a busy buildbot master, the simple stop() function in the init script can leave the sqlite locked and master will fail to start again due to sqlite operational errors. See http://trac.buildbot.net/ticket/2005 buildbot can't deal with this problem effectively so the init script has to be improved to make sure the sqlite transactions have finished before we shutdown the buildbot master process.
Do you have a proposed solution in mind?
No not yet. I just opened this bug to track the problem before I forget.
Is this still relavent with the latest buildbot releases. At my work we have moved to using a postgresql db backend for our buildbot installs. Before that, I had not noticed any problems with that in out buildbot-0.9 installs. If this is found to be still relevant, then please re-open.