Here's the response (from Michael Austin - C&D staff) to a post on the C&D forum when I asked the editors whether or not they measure acceleration times with rollout.
"Hi! We haven't changed our methodology, and we still use rollout. Currently (and going a long way back) we use 0.3 miles per hour to approximate the 1-foot rollout. The rollout time is typically around 0.2 to 0.3 second.
We've always had a lively discussion here in the office about publishing rollout times in the performance results. Do you think that would be a good idea, or just extra information that might confuse too many readers?"
I'm pretty sure MT does the same, but can't currently find a document that confirms that.