Regarding SATA and Xserve RAID performance, I guess it all depends on what you plan to do with the RAID. gekko513 has argued that the throughput bottleneck is not likely to be the ATA layer, which sounds reasonable.
But if you want to use the RAID on a file server or for transaction based systems, throughput is not all that matters. Simultaneous random access by many users may well be a more important issue in such a case.
And in this respect, SATA seems to have large advantages over Ultra ATA, at least if the SATA implementation supports Native Command Queuing (which is a fairly new SATA command protocol addition). Besides, the proponents of NCQ even claim that it reduces the mechanical wear and tear on drives. I believe that SATA with NCQ does have a noticeable performance benefit for multiuser random access based storage systems.
We had been putting a planned Xserve RAID purchase on hold for a couple of months now because we hoped that the next release might actually use SATA. But since it now does not seem to be likely that we will see SATA in Xserve RAID anytime soon, so we will go ahead and purchase the current version.