The current play end-to-end tests cannot be executed reliably as the publishers are sending their messages before any subscriber can be matched in a deterministic fashion. In order to push the repos further towards the ros2.repos integration, we disabled this test. See conversation here:
#104 (comment)

This can be re-enabled when a more deterministic start of replaying a rosbag is available. One way to work towards this is to introduce a --paused option to start a rosbag, prepare all publishers but don't start publishing directly. This allows discovery to match with the respective subscribers to guarantee receiving all sent messages.