Ryan,
I experimented with really long exposures this winter, and found that the 645D sensor is sensitive to what the astronomers call "spurious noise". The most noticeable artefacts look like curved or curled streaks, and are indirectly caused by cosmic radiation, no less, hitting the sensor. When rays hit during exposure, you get bright streaks. When rays hit during the dark-framing in "B" mode, you get dark streaks.
There's a science research paper describing spurious noise in CCD telescopes here (
PDF). Or if you take my word for it and can bear with my drivel, I wrote a blogpost on my find
here. :-)