Maybe they need to spin up a few dozen more VMs on EC2 to handle all of the couple thousand visitors a second HN is throwing at them.

It's more scalable, you see, you just need to wait until the wave has passed before the autoscaler is warmed up.