I think about this sometimes from a slightly different angle - not the "expected value" number of deaths, but rather the (utilitarian?) math of actions which af

I think about this sometimes from a slightly different angle - not the "expected value" number of deaths, but rather the (utilitarian?) math of actions which affect a huge number of people a little bit each. If you waste, say, 5 minutes each of a million people's time, in some sense that's almost 10 years of life lost. I think about this a lot when there are discussions about whether performance matters any more in software.