If it's peace of mind you're after, it seems like encouraging developers to target older versions of the Win32 API is a far more effective goal. Linux with Wine can have support for those forever, and Windows, with their almost pathological commitment to backwards compatibility will be able to run them them, too.
Linux APIs, in contrast, are far more varied, and change a lot more. So the "peace of mind" argument actually favors Win32.
> it seems like encouraging developers to target older versions of the Win32 API is a far more effective goal.
Perhaps, but I don't know how long that will last, and how hard that goal will be to migrate future windows to older windows. I don't see much point basing my goals on the uncertainties of proprietary software.
>and Windows, with their almost pathological commitment to backwards compatibility will be able to run them them, too.
Likewise, Windows does this for now. Microsoft isn't immune to changing course, and their track record isn't even great to begin with when we consider the 90's. I don't want to rely on the assumption that a trillion dollar corporation will always value legacy content. Enterprise tend to be pretty good at legacy support, but it still has a shelf life (unless you're COBOL I suppose. But I don't think Microsoft deals with as much safety/mission critical software as banks).
Linux APIs, in contrast, are far more varied, and change a lot more. So the "peace of mind" argument actually favors Win32.