How to Cure Disk/Controller (RAID) Problem?

[Jud]

I have an ASUS A7V333 motherboard with a Promise 20276 onboard chip
controlling two IBM 120GXP 40GB hard drives in RAID-0 (striped)
configuration.

This configuration has operated flawlessly for years.  However, in the
past two weeks, twice when rebooting after the preliminary stages of a
DragonFlyBSD installation (disklabel -B -r -w ar0s3 auto), my RAID array
has come up broken.  Rather than a two-disk array, it shows in the
Promise FastTrack BIOS as two striped "arrays" consisting of one 40-gig
drive each.  After putting the array back together using the BIOS
(deleting the two single-disk arrays and re-creating the two-disk
RAID-0), the first two slices/partitions in the array (Win2K and a
static pagefile) look normal, but the third gives nonsensical status
information in both Windows and *BSD, e.g., a size of 1600+ GB.  Neither
Win nor *BSD will format this slice/partition.

IBM's Drive Fitness Test (DFT) shows both disks operating normally. 
When I tried to use DFT to write zeroes to both disks preparatory to
reinstalling from backup images, DFT responded that this utility was
only available with IBM drives.  (??)  (This is an older version of DFT.
 I have downloaded the most recent version, but haven't had time to try
it yet.)

I'd very much appreciate advice/recommendations regarding where the
problem might be and how to cure it.

Jud