John Ament the existing test is merely a verification of a behavior strictly defined in CDI specification. E.g. that the controller cannot activate the context if already active and cannot de-activate it if 'something' else had activated it. I guess Tomas just picked Servlet as an easy and straightforward case to verify this. Therefore, I wouldn't use this package/test as an argument to place the tests you mentioned into CDI-TCK. In fact, I would also lean towards placing them in EE TCK as this is kind of spec-cooperation scenario.