If people need to use a particular older browser for a very narrow application, then they should keep an installation of that browser around for that specific purpose.
Its the same idea with modular programming methods. I build my js in modules. My comment system is separate from my login system. In the last couple of days I have been updating the Facebook Connect logic in my login system and I haven't had to worry about updating logic in my comment code.
They could ship IE9 with WebKit in addition to custom rendering modes for IE<=8, and consumers wouldn't even need to know. It'd save MS a lot of work, that's for sure.
Just think how hard it would push webkit/mozilla/chrome if they did manage to beat them on speed. It would push them to get even better. Compeition is always good as long as 2 or more projects stay alive.
You've just made an enemy out of every web-dev on HN :)
What is a compelling reason is revenue loss in their cash-cow franchises like Office. If they come to believe that they need a version of Office that works in the browser or they will lose revenue, then developing a fast version of IE is very important. If they believe that people won't use a computer that doesn't provide them with great experiences on the sites they love right out of the box, then developing a standards compliant version of IE is important.
I agree it's almost definitely technically infeasible.
I still can't believe they're maintaining two distinct HTML renderers...