> Thanks for any help. This worries me.
Seems like I was worried for good reason.

It looks like I had a corrupted FFS.

I ran an fsck -f after unmounting the fs and got lots of unknown file types,
unref files, duplicate inodes and partially allocated inodes.
I even had a internal fsck error, "dup ino with softdeps" or the like.

I've been using NetBSD for some ten years now and NEVER had any data
corruption on FFS before unless a disc failed.

This alll started with a mysterious

free inode /some/filesystem/<some number> had <large negative number> blocks

then

free inode /some/filesystem/<same number> had 32 blocks

repeating every minute.

The partition in question lives on a RAID. Shortly before the disaster began,
I had a timeout on one of the RAID components, thus an I/O error so the RAID
failed the component. This may be just a coincidence.

The unmount gave some error message "pending error" or so.

I'm running 4.0BETA2 (on amd64). And now I'm running a restore.

Any ideas what's going on? I'm not using UFS dirhash.