The rebuild of the world completed a while ago. No bizarre death. It
completed in 8 hours - three hours less than it took under 1.2. Must be the
extra buffer cache.

I think it's safe to say that the recommended changes work to mitigate
whatever was causing the filesystem corruption.

To recap:

1. get the new VM code.
2. remove the code in machdep.c that limits "nbuf" to 128.
3. config "options MAXBSIZE=3D32768" or "options MAXBSIZE=3D16384" (I used t=
he
latter).

Just two questions:

1. Will these work-arounds be committed to the source tree any time soon?
2. any progress on figuring out why it did this?

	thanks,

	Erik