What, exactly, are the symptoms of the MOD15_WRITE bug?

After spending much quality time with google, I still
haven't found what the symptoms of the MOD15_WRITE bug
are supposed to be.  I've seen an assertion that it only
happens with the Sil 3112 controller.  I've seen an
assertion that it only happens with very early SATA drives,
and an assertion that it does not happen with the Seagate
7200.8 drives.

I have a 3115 and 7200.8 and the write speed seems slow.
6.5 MB/s with write cache off, 7.7 MB/s with write cache on.
I tried changing dd to use bs=7k to avoid the mod15 but
it didn't help.  I have a new kernel standing by with
WD_QUIRK_SPLIT_MOD15_WRITE set waiting for a convenient
time to reboot, but I wonder if I'm chasing the wrong
problem?