Given the behavior expected in test 2 ("the first read in a causally consistent session must not send afterClusterTime to the server"), tests 6 and 7, which test for the absence of afterClusterTime in the first read of a session under varying circumstances, will always succeed.

It might be assumed that implementers will add a preceding read for this very reason, but I think it's better to be more explicit here and include that operation in the test prose. It's possible that drivers currently have implemented this test as written and aren't actually testing anything (I know I did...).