ONLINE LECTURES. Sure, traditional lectures are shown to be pretty ineffective. By why should the MOOCs be limited to "traditional lectures"? Interactivity can easily be built in -- the students have to type a number or answer a multiple choice "concept test" question or draw an arrow to show a specific force vector. It would be simple to incorporate these every few minutes and have the software provide extra "instruction" for those who give a wrong answer (with different feedback depending on the specific wrong answer).
MOOCs IN GENERAL. The model that springs to my mind is a single course, with "local experts". Pretty much every 1st year physics course follows roughly the same schedule (with a few variations). Why have 1000 professors each writing exams and deciding on appropriate homework and developing lectures and ... ? Why not have a "model" course that many schools share? Then much of the behind the scenes "grunt work" is taken care of, leaving the "local expert" as tutor and lab instructor and even "friend" (who is on the students' side against the faceless "course team" who get to play the part of the "heavy").
The exams could have multiple questions on each topic (analyzed to know which are easier and which are harder) so that "studying to the test" would not work, and not every exam would have to be taken at the same time on all campuses. (One potential problem would be developing a rubric so that each "local expert" graded consistently on exam questions or labs that require hand-grading.)
This also provides the students a better comparison of how they really stack up. Performance in the course would tell you not only how you compare to the 5 - 500 students in your own school, but also with students around the country. There would be no more "well, he got an A at a community college, but would he have cut it at MIT?"