So this is kind of cool. I didn't know this was happening but we've released a special Code Coverage version of the .NET Framework 2.0.
We have tools here that measure code coverage. Given a current scenario, what percentage of the code in a given assembly is touched by that scenario. It's primarily used by QA teams to judge their test automation to see how much of the product it's exercising. And we're looking at more ways of using it, since it's got the potential for being incredibly powerful. Imagine if there is a large bank of test suites of different varieties that are indexed by code coverage area. Developer makes changes to a given binary, you diff the resulting binary from the existing one, and then use that data to choose which tests with with to verify not only that the fix works but that it doesn't break anything else. Cool, huh?
So this version of the framework collects code coverage data (not sure what the performance penalties are but I'd guess there are some), and then pushes it back up to MS so we can see exactly what you're using in the framework, which will help us make better decisions around compatability and testing in general.