Opteron processors have Hypertransport links between CPU's, so memory latency to another CPU's memory is very low. Only a few cylcles. Contrast this with Xeon systems, which have to share their front-side bus for access to system RAM.
The real limitations would be with Windows, which doesn't support the NUMA memory architecture. This would prevent Windows from using memory installed on anything but CPU 0.
On all your other points, I hope he was just joking...