They've spun up "mega-supernodes," when cascading failures have taken the network down before[1], but they don't need to run them full-time, just when traffic is such that the traditional supernodes can't handle the traffic of a new network.

1: http://news.ycombinator.com/item?id=2048630