Also, the survey was of 100,000 Tweets - a drop in the bucket that must have some selection bias issues (even if unintentional).

9319 tweets/s at the moment

Cheers,
Scott.