As a practical matter this would most likely be a total non-starter - losing all that licensing revenue in one fell swoop would probably not be palatable to shareholders.

As a thought experiment, however, it is plenty enticing.

Vulnerabilities would be easier to find (at least at the outset), but with such a huge and slow-moving install base would probably be more serious.

I would imagine "translation" projects such as WINE and Mono would either massively accelerate development, or drastically drop off depending on how many *nix users would move to the new platform.