If you had been more cautious with respect to this one type of event at this one plant, you would have prevented this problem at a low cost. The problem is that only really tells you that 'if you know what's going to happen, you can do a really good job planning for it', which is rather obvious and completely useless.
My point is that if you were to have been more cautious in the overall specification for the plant (because, again, the plant was never designed to survive the type of event it was subjected to,) it is not clear whether the generators would have been moved (as that may not have been a high priority compared to other improvements), or even if they had, what the cost/benefit ratio would be. You have to remember that there may be many other events which may have appeared more likely than large earthquakes and tsunamis, and those would probably have been seen as more cost-effective safety enhancements, and would have been waste. In addition, this heightened standard would be applied to all the power plants subject to similar regulation, which would have been further waste.
The important question is not what the cost/benefit ratio for moving generators at Fukuhima before the tsunami would have been, but what the cost/benefit ratio for taking precautions against all events of similar likelihood and impact in would be for all power plants (or even all infrastructure) would be.
The NRC has had modernization plans at U.S. reactors that have spent much more than this. For instance the hydrogen bubble formation in that kind of BWR was well known here and reactors in the U.S. have had features added (dependent on a power source) that would get rid of it.
cost of preventing low probability events in the future >> cost of preventing events in hindsight
The problem is that you do not know which events will occur, so you will inevitably spend not only on implementing that $300000 upgrade on one plant which will need it, but on many $500000 upgrades on many plants (most of which will never be necessary or related to tsunamis). The cost/benefit ratio of preventing low probability events in the future should be studied, and only after looking at such studies can we decide whether any one change should be made. For example, the US government rejects any highway improvement which costs more than $3MM per life saved, as there are many lower-cost measures that save lives, and spending on less cost-effective measures is "statistical murder".
A study which I would find more important and useful than the one here would be a cost/benefit analysis of increasing all specifications relating to all (similar) low probability events. In other words, if you had been a designer of the plant looking to spend an extra $500000 to make the plant safer, would you have spent it on moving the generators? How many other contingencies would have been higher priorities than this one (if any)? What would be the total cost/benefit of applying this heightened standard to all new facilities? Would this be statistical murder?