Do you remember the rumors about only high end CPUs of the Nehalem architecture will be overclockable, that was spread a while ago? Just a wild guess, but I have the feeling the 4 GHz QPI limit is some kind of a purposely put limit. I am having a hard time to believe 4 CPUs on 4 completely different boards fail at about the same BCLK limit. Maybe I am doing something seriously wrong or it is no coincidence (low end OC limit?) =)

With the Biostar X58 I can boot right at 222 MHz BCLK, run 8M (didn't try more, Prime crashed) and using SetFSB anything above 222 MHz is not even stable in idle. The only thing with a similar behaviour were low end C2D chips with really hard FSB walls. Cold helped to push that wall further, which seems not to be the case for Nehalem CPUs.

Just a wild guess, but I would love to hear some scientific explanation for this "issue" that almost everybody seems to face.