Every time someone in the nuclear establishment says that a particular kind of horrible worse case accident can't happen, there is a one word answer: Fukushima.
This. Also, part of the technique for the 'can't happen' brush-off is to quote enormous odds against. After the 2011 Japan earthquake/tsunami we heard first how such things were about one-in-a-thousand-years, now we're hearing The NRC staff in its 2014 study said a major earthquake could be expected to strike an area where spent fuel is stored in a pool once in 10 million years or less.
They have all omitted to mention the 1896 Sanriku earthquake and tsunami, which was practically as devastating as the 2011 event, and in the same general area, killing even more thousands of people. Maybe the experts would try to use the excuse that the magnitude in 1896 was just a little smaller than in 2011. But it certainly was in a similar league, and so the risk of such events causing that order of devastation in that area is more like once in 120 years, not 1000 or 10,000,000.
Ok, so that applies to a specific seismically active geographical area. But the unjustified brush-off merchants are mobile, and express their 'expert' views everywhere. So we need to beware so-called experts using brush-off statistics, and look carefully into their so-called facts.