I've been pretty highly entertained by the complete and utter inability of the other commenters to recognize her examples as edge cases.
Mayer's explanation should really be read like this:
Employee X has been working 130 hour weeks for the past month. He's looking more than a little worse for wear; he's probably going to burn out soon. Let's sit him down for a chat and make sure he's doing okay. Oh, he's been missing his daughter's soccer games to manage a critical deploy, and those soccer games are super important to him. Alright, let's deal with that. Mandate that someone else take over, de-prioritize the deploy, rework the process so that deploys happen on a different day or a different time... whatever it takes. He might still be working 130 a week, and that's not great, but he gets the critical thing he just can't miss.
"When Google was a young company, she worked 130 hours per week and often slept at her desk." - Nothing from the article suggests this is an edge case. Instead the author seems to perpetuate a very dangerous meme "toughen up and work insane hours if you want to get anywhere in the startup scene".
This meme is dangerous not just because of the effect it has on people who don't subscribe to this, but also because it is just plain wrong. People who work longer hours are not more productive and they do not get more done. Its a simple case of measuring input (hours spent) instead of output (quality code, decisions whatever...), because measuring what matters is not very straight-forward.
I know the statement above doesn't have backing data, just my experience/ observations, but have you seen anything to contradict this statement?
Mayer's explanation should really be read like this:
Employee X has been working 130 hour weeks for the past month. He's looking more than a little worse for wear; he's probably going to burn out soon. Let's sit him down for a chat and make sure he's doing okay. Oh, he's been missing his daughter's soccer games to manage a critical deploy, and those soccer games are super important to him. Alright, let's deal with that. Mandate that someone else take over, de-prioritize the deploy, rework the process so that deploys happen on a different day or a different time... whatever it takes. He might still be working 130 a week, and that's not great, but he gets the critical thing he just can't miss.