I can't find the article now, but I remember reading that the IE7 mode in IE8 had a ton of little gotchas. They were pretty much all the type of things that weren't a big deal on their own, but cumulatively it mad testing for IE7 on IE8 totally worthless. I think the tests would be unreliable, plus the amount of minutia you would have to mentally track about what is emulated properly and what is not was huge. My feeling after reading this article was that the simplest route, unfortunately is to have virtual Machines with the different browsers on them.

Basically to provide 100% pixel perfect on IE(6,7,8,9) suupport:

IE versions can't be installed side-by-side (and the un/install process on Windows is very inefficient). So you have to have to have at least 4 VMs/boot partitions (at least 1 of which must be running XP), each loaded with a different version of IE. Each with a different version of debug tools.

And unless you force (strict mode) on IE8(& 9?), test IE8 in IE7 mode, IE8 Compatibility mode, and IE8 Strict mode....

Fortunately my employer dropped IE6 support 2 years ago. Plus, I find that usually just testing for IE7 in IE8's IE7 mode, and forcing IE8 to render pages in IE8 strict mode, is generally sufficient 95% of the time.

How faithful is that rendering? The IE7 emulation in IE8 is not an identical reproduction of how IE7 itself renders pages.

So we should probably be testing IE6, IE7, IE7 emulated under IE8, IE8, all of the other ones emulated in IE9 and IE9.

Ouch!

If I remember correctly you can force IE8 to render "on edge", basically avoiding the compatibility mode. I'm guessing you can do the same thing for IE9, dropping all these tests to just one for each browser.