I have never heard, and I always just assumed they rotated them. It's a good question, but I dunno if we'll get it answered unless we get an ASA again.

I don't use it; I serve up alternate ad styles via PHP and assign a channel to each one, and that's how I control performance (why won't anyone click on green? it's my favorite color)

Yep. Indeed, I suppose I really should do some A/B testing, but I generally prefer to just let them get on with it.

If you have three color palettes rotating, say red, green and blue, the only way to see which performed the best was to have the same ad shown every time someone viewed the page - otherwise it may be different ads that are causing clicks, not colors.

In my experience viewing my own sites, the colors seem to change on a planned basis but the ads are more random. That's led me to assume (a dangerous thing) the colors change based on repetition not based on which ad is performing best.

FarmBoy

Seems random, see [google.com...]

"You can select up to 4 palettes that will rotate randomly."