The problem with simply displaying 3D models is that it is not flexible. Right now, there are working ports of Quake to WebGL, you wouldn't be able to do that with if you limited yourself to providing a default engine.
I think the major turning point could be when the dominating mobile browsers (read: webkit) adopt WebGL and have decent performance, I think things could change. The original iPhone's view was "all apps are on the web", and now comes a bit closer. I don't believe these would actually replace native apps any time soon, certainly since both Apple and Google now invested too much in their app-stores and native development kit, but a lot of these "native" apps already are simple embedded web-views wrapped in a tiny native application, and this could then easily expand to games.
The issue here at hand is, the major market for 3D is games. You won't see a fully-featured professional 3D modelling/rendering application any time soon as web-apps, even if WebGL would be succesfull. Games right now, are dominated by Microsoft's DirectX, on which they bet big, and actually won (for the time being). They dominated OpenGL for the last decade with only a few exceptions (the Quake engine being one) and will be reluctant to adopt WebGL. Right now however, Valve is betting on Linux and Mac, and the latter in general is gaining grounds in normal households, which means these are OpenGL-only machines. Yes game devs can use Wine libraries to translate DX calls to OpenGL, but this always has its downsides. Also, all mobile development, where the majority of new games are developed, is dominated by OpenGL (ES). Point is, from now on, DirectX will only lose ground. Maybe some day, Microsoft will be forced to adopt WebGL..