And this fits the data: why would companies not dramatically push people to work fewer and fewer hours for this magical increased productivity that you're claiming? They could market it as a perk AND get so much more done!
What I think is happening: there's little cost to companies to encourage workers to work 60 hours instead of 40. Sure, they won't get 50% more done, maybe only 20% or 30%, but the company doesn't pay overtime, so that's a free 20% or 30%.
Of course, there's some point at which overwork does start to impact overall productivity vs standard-40-hours, but it's much, much higher than 40 hours. This study  seems to indicate that little extra gets done from 55-70 hours, but from my reading, that still doesn't indicate that someone working 70 hours gets less done than someone working 40 hours. Looking at figure 4 in that study, it looks like you'd have to go to 90-100 hours to see output drop back to 40-hour levels.
It just seems ludicrous to me to suggest that for the population at large, working 40 hours is the optimal amount, such that anything less will result in greater overall productivity, and anything more will lower overall productivity. Really?
Currently leading a team I see different behavior from team members working longer hours. For example, we had a conversion to do that was running into a deadline. Together with a team of 4 we worked into the evenings to get it done. 2 of them 'compensated' by being there but browsing more Reddit, the other 2 actually sat down and got the work done.
I see a lot of comments justifying the first kind of behavior, and it doesn't make sense to me. If you are not productive because you are on HN, you're doing that to yourself. At 40-60h we're in the realm of that being a choice, and then you're choosing not to be productive.
FWIW, I'm not trying to justify people working long just to show that they are there/committed. That's a terrible practice too.
In software development, I'd argue there's rarely a significant difference between marginal and overall productivity. It really boils down to personalities however, let's be real; and I am not gonna claim my point of view is universally true because it too isn't.
The more and more you work, the more BS you start to produce. You are tired, you start to imagine your bed (or girlfriend, or TV, or whatever floats your boat) and you start putting less thought into what you do and just go into "spear mode". One of my former managers used to say "If I see you tired, I prefer letting you go at Friday 14:00 so you don't spend your entire Monday fixing the bugs you introduced during the second half of that Friday".
This is a real phenomena. Granted it doesn't apply for everyone, but IMO you discard way too quickly those of us that do need some freedom and some "slack" to actually be more productive. And it's very generous to call the normal legal requirements not to work over 40 hours a week "slack".
This is also real: when I am having a very hard time and I can't solve something and I get nervous and stressed out, and if you're my manager and tell me "Dimi, just call it a day right here, you can do it tomorrow" -- you can be VERY sure I'll do it tomorrow. While conversely if you tap me on the shoulder 10 times for the remainder of the day, I can guarantee you I'll get absolutely nothing done. Different strokes for different people.
So no, "the free 20-30%" isn't something that the company can get from everyone.
Or are you actually in agreement about the shape of the productivity vs hours worked curve but you disagree where the peak is for software developers? If so, where is that peak do you think?
It's the latter from your comment -- I disagree where the peak is for software developers. But I can't give you absolute numbers that I can claim are true for everyone. There's no such thing. I am a proponent of the per-individual approach.
That being said, if we try and devise a good median number -- I've read quite a few articles several years ago that claimed that 27-32 hours of work are not only the peak of the programmer's productivity but that's also the working schedule that makes people want to work more in their free time. So in terms of your "free 20-30%", for people like myself a 30-hour work week would likely result in free 50% once every several weeks.
Not feeling forced into unproductive schedules gives your brain and psyche a good leisure time during which you occasionally find yourself thinking about the work problems in a chill state -- and it's also often the case you reach for your laptop on the couch to try and solve them outside of the working hours.
Again, it's subjective. I knew several people who worked best when they were given tight schedules and impossible work hours. I have to somewhat gloatingly add that all of them ended up in hospitals and subsequently became very strong advocates of low-to-medium working hours. So IMO the "punch the sun!" people don't live a sustainable work life.
Now in this situation one can try to get that bump again by going to 60 instead of 50. Or one can reduce hours from 50 to 40 and suffer (in the short term) a negative bump with the idea that it will bounce back. Given that the short term bump is immediate, and the readjustment takes some time to make itself shown; and given that short term results dominate ones thinking, we can see how one will continue to choose longer and longer hours.