What's stopping us from making a tool that parses HTML/CSS and points out constructs that are known to be bad in certain browsers? Sort of like lint: "It looks like you're sending element foo inside element bar to IE6. This probably won't work right"
I personally don't have a windows box to test IE, and I'm not looking forward to the day when I'll have to test on IE. Surely there is a better solution than manually testing browsers?