Aye, I understand that. Eventually a fiscal decision was made, rationally, by a company teetering on the edge. And that's OK.
But speaking as a developer, I keep going back to my earlier statement; each moment when you're working on code and you have to make a decision, you weigh up the pros and cons of each option and pick the one that you want. The decisions they were making back then were, each time, to choose a windows-specific choice.
Somewhere along the way, the small marginal improvement in development time outweighed the benefits of keeping the code portable. I'd understand if it was a pathologically windows-only tool to begin with, but it wasn't; it was a fantastically portable codebase that they chose to decimate!
I would wager that when they eventually developed mac support, the time and developer resources that it took was way more than the time saved by choosing Windows-only options in the past. As someone who works on portable codebases a lot, the ability to run Valgrind on Linux, and Shark on OSX, is alone worth the extra time it took, because it so significantly offsets debugging time on windows. Difficult-to-debug bugs that manifest only in rare cases on one platform oftentimes manifest far more easily on other platforms, somehow that's the nature of the beast.