Using "biggest in recorded history" as an approximation for "biggest possible" seems weak to me. Proper engineering will look at the numbers and account for events outside of the historically observed range, but within the probability distribution. The "biggest event we see in 1,000,000 simulations of the lifetime of the plant" is likely much larger than the "biggest we've seen in history."

Which isn't to say that it was bad engineering. Just that this being the biggest quake in history is not an excuse.