> So that's at most a few thousand deaths for something like 100 Million reactor hours (napkin math, don't cite).
Sounds about right, but when you factor in current designs and requirements it's 100x better.
LRF (large release frequencies) are required to be less than 1 in 1 000 000 per year of reactor operation [1], so roughly 10 billion reactor hours. The calculation of this considers at least 1 in 10 000 year levels of high winds and earthquakes [2].
Since the typical reactor is 1GW, so ~8TWh per year, and global energy consumption is ~23 000 TWh per year, we could use roughly 3000 reactors globally to meet most of our needs, which would put an event like this roughly every 100-1000 years.
[1] https://www.nrc.gov/docs/ML0909/ML090910608.pdf
[2] https://archive.opg.com/pdf_archive/Nuclear%20Licencing%20Do...