the "dup alloc" file system panic you reported occurs when:

	- ffs wants to allocate a new inode (for file creation)
	- it picked a cylinder group with free inodes
	- it looked in the allocation bitmap for the CG, and found a free one
	- it then looked at that inode, and discovered that it wasn't actually
	  free (the file type field of the inode was not zero)

now, this could happen if the allocation bitmaps weren't up to date;
this is one of the things that fsck is supposed to repair.

In practice, I've seen this happen on BSD's (since the 4.3 era) when
you mount a "damaged" filesystem after a crash w/o fsck'ing it first.

Disk corruption could also be to blame.  (I could speculate on various
ways in which hyperintelligent disks might "helpfully" reorder writes
close to a shutdown/power off to cause this sort of corruption to go
undetected..).

					- Bill