So, what they’re basically doing is loading straight from long-term storage into each core’s memory cache, and just running a crapton of cores? That’s actually a pretty good idea provided your long-term storage access speed is fast enough.