Matt Lebofsky posted long ago that many of those were caused by the hardware blanking track not having quite enough data to do the blanking at the end of the channel. In that case, the "error" is trivial because most of the channel was turned into WUs before the error happened.
But of course the software blanking is derived from the data so shouldn't run out too soon. I suppose careful study of the AP splitter code would yield a list of things which can cause it to error out, but wouldn't get us any closer to being able to guess which are most common.
In short, I dunno.
Joe
