I agree. After all, developing behaviors for robots is the Holy Grail of mobile robotics. The development of robotic behavior is still a very active field of research, although it is a problem in which many people have been working for more than 40 years: reactive, hybrid, three-level architectures, sumbsuption, … All of them try to solve the same problem: generate apparently intelligent behaviors.
ROS provides a great environment to develop software for robots, but it does not implement the generation of behaviors for complex tasks. Until recently, Smach and ROSPlan were the only outstanding projects to address this problem.
I think it is positive that many of the approaches that have been developed over these years are released to the community as ROS packages. It is a way for these software not to die with the project that saw them born. As they say in the YARP paper, sometimes projects are black holes of code, of which only papers survive.