40 hours is the peak of the Laffer curve [productivity being inversely proportional to hours worked a week]. I have absolutely no data on variance for the peak of the Laffer curve (if you imagine attempting to gather this data, you'll notice it is extremely difficult). That being said, despite this lack of data, I have heard a lot of people claiming they can work a whole lot more and get more and more done. I do not believe their ad hoc methods have secretly outsmarted science. It seems more likely that they are mistaken (if only for the reason that it is very hard to measure and they universally lack specific reason to choose "my peak productivity happens at >60 hours a week" over "the volume of my work is more noticeable than the quality of my work", which is essentially universal -- but lines of code, as a metric, don't pay the bills).
Of course your point still stands, they'd just have to be even more of an outlier.
You wake up, you know that your limo is waiting in 20 minutes. You'll start with a scheduled call on the way to gym. At the gym personal trainer makes sure you're giving maximum. Then the limo takes you to the private jet so you can catch the conference in Europe at 4pm. etc.
My point is, if the world around you revolves in a rhythm where falling out is not an option, you can find the balanced flow and rhythm so 100+ hour week full of balanced work is absolutely normal. You just stop wasting non-work time over the weekend and use it for maximum recovery and family-time.
Add a bit of humour and laughter to these 100+ hours and stop caring about issues you can't fix and it's a paradise to live in.
However, that's a big SHOULD/COULD/WOULD. Real world statistics probably reflect that most people either aren't capable or simply wouldn't maintain this kind of routine.
As opposed to "your project is behind, time for 11½ hour work days." I got that earlier this year. At least I was paid hourly at the time.
I doubt that the people who work[ed] on saturdays had a much different daily routine than those who didn't; but still, working that extra day only results in less work done.
You work that extra hour/day, week after week, and you see real productivity gains (not necessarily per-hour productivity gains, but marginally speaking, you make progress faster). As others have mentioned before, this trick can work very well in the short term.
However three months into it (roughly +150 man-hrs), you make a small blunder that nobody notices at first. One year later, it will blow up and set your whole team back for a whole month (assuming team_size = 5, -800 man-hrs). Everybody will try to catch up by staying an hour later, or working on the weekends... rinse and repeat.
The problem is that cause and effect are so far apart in time that nobody notices why this is happening. Usual suspects such as "rusty codebase" or "technical debt" get assigned with the blame and no lesson is learned.
It's NOT a tradeoff of more volume for less quality and increased risk. All the drawbacks you list are valid, and come on top of less total productivity.
You don't get a product built faster but with more defects and technical debt. You get the product built slower; with more defects and techical debt; and at a cost to the workers personal life in a true lose-lose tradeoff.
* WWI = death by explosive * over work == more defaults => more of your soldiers killed less of theirs, more money spent * so british army edicted a ban on overwork ~ 1917 followed by USA.
The question is does it matter more to produce the most defective products, or to produce stuff that works?