Doesn't this destroy existing "free" pages that have data in them that might be re-used later? If it's based on what people run, wouldn't that data already be loaded from the last exit? This may be good for initial start-up, when the memory is free, but then drive usage is pretty high during that time frame anyway.
I'd be interested to know if this actually speeds anything up.
Yes, that's basically what it does. I came up with some negative conclusions like yours after I finished experimenting with it. You can read my findings here: