Given that the issue regressed between kernel versions 2.6.32-20 and 2.6.32-21 that really suggests it's a kernel bug, not X, especially since you see the corruption so early in the boot process (maybe before X has even started).
Like I mentioned, I would upstream this however you have to test the latest development version (maverick) first. If you are not interested in doing this, and just wish to get your system running, I would recommend using the technical support channels.
Given that the issue regressed between kernel versions 2.6.32-20 and 2.6.32-21 that really suggests it's a kernel bug, not X, especially since you see the corruption so early in the boot process (maybe before X has even started).
Like I mentioned, I would upstream this however you have to test the latest development version (maverick) first. If you are not interested in doing this, and just wish to get your system running, I would recommend using the technical support channels.