I rebooted a system and got error messages something like this:
Superblock has a bad ext3 journal (inode 8).
*** ext3 journal has been deleted - filesystem is now ext2 only ***
Superblock doesn't have has_journal flag, but has ext3 journal inode.
The system has a large external RAID, which is too big for one SCSI LUN on the given hardware. So, the data is sliced into 3 LUNs, joined together with RAID0 using whole partitions.
I think that the above problem happened because the fstab has "LABEL=name" instead of "/dev/md0". I now realize that by concatenating whole partitions, the first LUN has a valid label, and the system tried to mount the single LUN instead of the assembled RAID0 array. This seems a bit obvious now, even though it has been rebooted several times with no problems. I think it can be recovered by fixing the RAID0 config.
The lesson is to avoid RAID0 with whole partitions and using LABEL=. But, I also think that this configuration should give some warnings, which it didn't. Shouldn't the system give warnings when more than one block device has the same filesystem label? This system is running FC6, so it is possible that this has already been addressed.