Grr.. I spent most of the day trying to track down a bug that was appearing in my Python interface to BerkeleyDB 3.0 causing the python process coredump (*gasp*)! I believe I finally tracked it down to a bug in the BerkeleyDB 3.0.55 library when aborting a transaction while using subdatabases, it'll be interesting to hear back from them on this one. I sent them a simply example in C that reproduces it on my system.
For anyone interested, I'll be releasing my Python BerlekeyDB 3.0 interface very soon and will post a note here as well as to sleepycat and on freshmeat.
PS if anyone actually reads these, A certification would be cool.