Roy has a very interesting post on the unit tests that we shipped.  This is the kind of feedback we want.  It is a very interesting idea about how we only enable the tests that people can run.

IMHO, the deficit is in the tools.  How do I know what tests to disable or not install?  In V2 we are going to separate out the tests (of course this means we have to make some of the classes / methods / properties public that we might not normally, but hey, it is all in the name of testing right ?   How would you expect the test to be fully runnable?  Should we control this through some configuration?  Create suites of tests?  Gen the tests?  I really would love some ideas.