Long article, but the best one I've run across explaining the inputs used to choose design tolerances, the advances in those methods over the years, and where the Fukushima design and later maintenance was and wasn't in accordance with best practices.

The overall conclusion is that it was known by the early 2000s that the plant's worst-case tsunami estimate, used as its design basis, was no longer sufficient, given improved understanding of tsunamis. At that time, a significant upgrade should have been initiated, since the original design-basis tsunami of 3.1 meters was no longer believed to be a good estimate of the worst case. But those reports basically got buried and nothing was acted on.

Side note: I was wary of submitting, because we've just had a big round of Fukushima and nuclear discussion/debate, and rehashing the same debate in another thread doesn't seem productive. But imo this article is quite informative and worth a read.