Doesn't this assume that this cost for commercial OSes is zero? In my experience, that hasn't been the case (as an OSX user, haven't used Windows for software development in a long time).
Just as an example, at our company, we run a Windows Server 2008 R2 Terminal server, with Outlook 2003 installed for email. A while ago we decided to replace the aging Internet Explorer 8 installed on that machine (shudder) with IE11 (still shudder - but it's company policy), and if people are going to use IE, I thought at least they should use a reasonably current one.
Turns out, that Outlook uses IE for rendering HTML emails, and that Outlook 2003 using IE 9 or newer for that purpose results in HTML mails being truncated randomly. Since some of our users run Outlook almost exclusively in their RDP-sessions instead of their local desktops, this was a rather big deal for us, and it took me a while to figure this problem out.
And keep in mind, this problem was caused by the interaction of two problems written by Microsoft itself. Imagine the non-deterministic fun you can run into with applications by a dozen or more software vendors, running across several versions of Windows, 32 and 64 bit.
(Yes, I kind of needed to vent. Sorry.)
Now that Office 2003's support has ended, I don't think Mirosoft will do anything about the problem for newer versions of IE.
(If I could, I would also upgrade Outlook on that terminal server, but that is another story altogether.)
Also, as an ISV or IHV, if your app or device fails to work with Windows, it's a DEFCON-1 emergency. If it fails to work with one of 5,000 random obscure Linux distros, well, they'll get back to you on that, someday, maybe. It's simply economics in action.
Note that neither of these are issues in the tablet or phone markets, which is why people citing those markets to defend Linux on the desktop are all wet. (They're in good company, of course -- lots of people in the Windows world have failed to understand why those markets are different, including the developer of Windows itself.)
As an open source fan but also someone who has been burned by the hidden costs of implementing it in certain emerging market regions (extensive training, time, lack of people to fix it when it goes wrong, incompatibilities etc etc vs Windows environment with a wider base of support and usage), I would find this very interesting.
I think things often look different in London or SF but the idea of "free software" in poorer parts of the world is often a false one.
FWIW OpenBSD, which I had installed on another partition, ran flawlessly with the exception of suspend/resume. This was a couple of years ago, I think there have been improvements since then.