[skolsuper]

> Only no-SLA consumer apps. Just having a price tag on unavailability will change most of the reasoning for his points.

I don't think so. The gist of the reasoning is that it's better to have more numerous small failures that are fixed quickly than rare but catastrophic failures. I think that's true even when you have an SLA, which is probably why it's broadly in line with the principles in Google's Site Reliability Engineering book

Others have made the point though that this all goes out of the window when talking about safety-critical engineering (Therac-25, 737Max etc)