"The advantage of buying, say, September VIX puts is that the huge 6.87 point differential between VIX and the Sept VIX futures will have to disappear by September VIX expiration (in eight weeks). Remember that the Sept VIX puts are priced off of the futures, not off of VIX. So, if that 6.87 differential shrinks by the September futures falling from their current price near 31 to the current VIX price near 24, those puts would profit handsomely."
He's right that the future is most likely to decline and meet the index than other way around, but his explanation is ambiguous. Considering mean-reversion effect by itself we should actually expect the opposite - VIX index wobbles around all over the place but eventually comes back to some steady level. So in our case we should expect VIX to rise up to the futures level. This is a valid hypothesis, however most of the time VIX futures display an upward slope, so mean reversion cannot be the only factor.
The explanation why VIX futures are most of the time higher than the VIX, is in convexity premium. The distribution of VIX levels is very right-skewed, and that skew demands premium. Every day that premium is getting smaller, and futures get cheaper. Meanwhile if VIX is below some level it will move up, if it is above that level it will move down. This interplay between mean-reversion and risk is what drives the complicated dynamics of VIX futures.
When I have more time I will try to post some charts, and more quantitative description of the dynamics.
On a separate note here is a very thorough VIX introduction from Mr McMillan.