I have argued this as well. The key really is the facilitation of student interaction, and that requires a savvy guide in my experience. I've seen it work with biology, with botany, with research writing, with new media studies in a seminar format that was highly "traditional" in that we read a text, met, and talked about it, etc. For me, critical mass occurs with students blogging in their own spaces and a "motherblog" that aggregates those blogs. I wrote about some of my ideas in this post: http://www.gardnercampbell.net/blog1/?p=620
I also think Carl Rogers' notion of "freedom to learn" gets at what Howard advocates and what I advocate as well. Rogers talks about this principle in terms of a science class in one chapter, anticipating the STEM argument that student-to-student interaction won't work in what's often called a "content-driven" class.
I think where Jim and I may differ is on the value and nature of study--at least, study in terms of diving deeply into long texts. We haven't talked about this for a few years. When it came up last, "edupunk" was brewing. Jim may feel a little differently now.
I do think we need more examples of connected courses than the three highlighted in this course. ds106 in particular is highly performative and its many tropes and practices may not be generalizable into other areas, though it's interesting to see how strong its pull is in terms of how people are interacting in this meta-course. Nothing wrong with that, but anything with that strong a cultural pull tends to skew the kind of activities and investments people commit to.