A post at Vix and More includes the following graph of the VIX and the forward-looking 21-day realized volatility:
The post discusses the fact that realized vol has remained well below implied vol, but I think there's a much more interesting facet to this chart.
First, consider what is being plotted: the VIX, marked in red, represents (in theory) the expected volatility over the next month. The green line shows the volatility that will be realized over the next month. Thus, if the VIX were actually representative of the future, the two lines should overlap. But they don't!
Instead - and I think this is the more interesting bit - they lag. Shift the green line over by 21 days and the two lines are a near-perfect match. Such a graph, instead of lining up future volatility with expectations of volatility, would show realized volatility against expectations of volatility. The fact that the two line up so well means - brace yourselves - that the VIX is much more reactive than predictive!
In September 2008, the VIX spiked higher. Note the green line, which spikes higher as well - 21 days earlier. In other words, they are both reacting to the same market events. This shouldn't be surprising to anyone who believes in remotely (or occasionally) efficient markets, but it's a good thing to keep in mind: the VIX in practice is more a measure of realized vol than anything else.
To tie this back to the point on VIX and More: in hindsight, of course you should sell (delta-hedged) options into a sustained rally; the reduced vol in the future will be much lower than the reactive vol of the past which the VIX reflects. This is simply the lagging effect shown above. But this isn't a function of any VIX magic, it's because rallies tend to be more gentle (lower vol) and drops tend to happen more violently (higher vol).