to be fair, they needed to scale to roughly a thousand concurrent users (where concurrent is used very loosely here). so I wouldn't say that scaleability was really a requirement. 1000 users is nothing.