We chatted with John Poole of Primate Labs, who highlighted the substantial improvements in many single-core measures and in memory performance, suggesting that lower multi-core scores later in the Integer Performance testing run could be indicative of thermal issues.
If this is true, and they can't fix it, which could be difficult if it's a fundamental design issue, then it's a serious problem. Essentially what he's saying is, the test scores didn't hold up as you might expect as the different multicore integer tests ran, suggesting that when the machine is being hammered it starts to overheat and the processor throttles itself back to protect the hardware - hence the scores start to dip.
I hope this proves to be unfounded, otherwise I'd be tempted to go down the 2012 refurb route with that nice roomy case for expansion options.