The issue is that the example nostromo gave (40m) was not intended to be predictive of what would actually happen. It was based on a worst case / left unchecked scenario (useful for establishing an upper bound), and therefore irrelevant w/r/t the system responding to the model.