Could it be that those failing are failing because their ideas are just not as great as they had imagined?
I can do PCB routing, design and layout till 5am in the morning, but the day after I need to sleep well to do more conceptual work, or else I'll forget what I'm working at.
[0] - as long as it's not Work. I'm a sad case of extreme intrinsic vs. extrinsic motivation issues; my best well-rested extrisincly motivated hours are about 25% as productive as sleep-deprived intrisinc-motivated ones. I have no clue how to fix this.
John Carmack can pull insane hours and retain productivity. Also, Usain Bolt can run hella fast. If you think that anybody can run as fast as Usain Bolt if they just suck it up and try hard, you're going to be very disappointed.
The root of all of this (IMHO) is that our managers are currently incompetent. Good managers, motivate people, get their people to deliver on-time, and adapt the environment to make all of this happen. There is a difference between managing an organization ("You can't manage anything you can't measure") and managing people. The latter, I believe, is the problem.
It all comes down to the managers, who, unfortunately, are often influenced by this "longer hours must be better" mentality which is simply not true.
But as a project wears on new code more often than not is to repair broken aspects and not so much to build the new and exciting features. Constantly switching gears between different parts of the code base can be hard to maintain that beloved flow.
People are tired and start making more and more mistakes or less then optimal decisions. Or they spend more time chatting with colleges or staring into screen without doing much or do longer meetings cause they are only option for socialization.