I agree on one level (literally, not figuratively) - if you have awesome (or even bearable :P) deployment for your software stack. We can do that. On the other

I agree on one level (literally, not figuratively) - if you have awesome (or even bearable :P) deployment for your software stack. We can do that. On the other level, there's stuff we see as core infrastructure (for example Hardware, or even some parts of OpenStack running on that hardware) - of course there are also downtimes and emergencies and dumpster fires - but they are pretty much unique little snowflakes and the repeating ones happen a few times per year. There simply is nothing "to deploy". Maybe one can argue that "runbook" is not 100% correct, sometimes it's a runbook including debug info. But it's not turtles all the way down, and I stand to my point there's stuff where the cost to benefit ratio totally ends up at "automate it away" and there's other stuff where it's the opposite.