The outage was caused by overloaded supernodes, from the article it is hinted that less than one percent of the Skype clients act as supernodes. If supernodes are this prone to get overloaded, why did Skype not use more clients as supernodes?
The article says supernodes help to establish connections between regular nodes. Does that mean the supernodes are responsible for NAT hole punching?
Would the outage still have happened if none of the Skype clients were behind NAT? Or would a situation where all Skype clients had a direct Internet connection have meant less load on supernodes due to lack of need for hole punching and more nodes available to act as supernodes?