the idea that rises in popularity can be important even absent specific contagion is known as threshold models and significantly predates Watts' recent work discounting the effect of hubs. Granovetter's 1978 AJS article lays out the basic idea which is that actors can be sensitive to the general level of prior adoption without specific dyadic links being important. different version under this general family of theories include Schelling's tipping point model of segregation, David's network externality model of product adoption, and the Bikhchandani/Banerjee model of information cascades. the best lit review that i'm aware of that summarizes and categorizes the various versions is in Valente's book Network Models of the Diffusion of Innovation.

also, while it's true that Watts was trained in physics, it's a mistake to think that we social scientists merely ape the hard sciences. with the exception of a few bridging figures like Watts, the social science and physics literatures on diffusion and networks are largely independent and the social science findings often anticipate the hard sciences by decades. for instance, in his 2003 book on preferential attachment Barabasi (a physicist) writes that Google has achieved lower market share in search engines than Microsoft Office in productivity suites and leaves it open as to why -- when econ and soc have known since 1985 that the explanation is network externalities.

Threshold models also explain a very important characteristic of innovation, that of parallelism. The telephone, the airplane, radio, television, and numerous other groundbreaking inventions were developed simultaneously by several independent inventors. Had Einstein not published his Special Theory of Relativity in 1905, half a dozen others would have done so within a few months time.

It's not merely whether and how one innovation is adopted widely, it's also how the innovations derive.

Which is where the articles simplistic assessment of "everyone is connected to everyone" falls flat. Even in the highly connected internet world, "friends lists" still provide propagation routes. Part of what makes the "threshold" possible is that enough credible "connectors" have been spreading the idea that this MIGHT be doable and therby preparing the audience.

Note also, that neither of these explains how the epiphenomenon usages that are unintended in the original invention, flow. These two are largely "threshold model" based disseminations.