And yes, we know that JS is interpreted and that it's no use for building anything more than a clock on your webpage, right?
WebGL is just that, a poor-man's subset of OpenGL. As is OpenGL ES 2, which is a standard that's deprecating very fast as mobile GPU's pack more and more kick-ass.
OpenGL ES standardization efforts are badly left behind the development, as graphics vendors are adding more and more extensions to GLES in their implementations. These extensions tend to be a little stripped down versions of desktop GL features and the worst thing is that they're not quite compatible with desktop GL or even similar extensions form other vendors. Next version of GLES (codename "Halti") has to be standardized pretty fast or it will be out of date by the time it comes out.
If you want to write high quality 3d games or graphics using GLES2, you will have to use vendor specific extensions and write code specifically for the devices you intend to ship on. If you're doing something very basic, you may get away with using only the features in GLES2, but you'll still have to consider how to match the performance to the wide array of GLES2 conformant devices out there. It also helps if you're a part of a major graphics vendor's developer partnership program, like it or not.
WebGL's standardization cycle probably even slower. It took a while for the spec to reach maturity and major browsers implementing it. AFAIK, there's no extension mechanism to gain access to not-yet-standardized features. When the next version of OpenGL ES comes out, how long will it take for it's new features to come available to WebGL? Is the WebGL API even designed for extendability? Who are the industry players willing to contribute their engineering efforts to building WebGL specs and implementations?
As for your second question, JS is very useful, I love JS, I write 90% of my code in JS. But we're talking sophisticated 3d apps here, not whether it's useful at all.
http://www.youtube.com/watch?v=DQrC5YLKFUY
http://www.youtube.com/watch?v=Lmvj6-CFbbA
JS is not very pleasant to code with, so writing an actual game logic could be a bit painful, but so is development for PS3 and it's done somehow. It's not a question of performance, rather about workflow and structures provided by language.
Though, I think I've heard pretty much all the complaints against JavaScript, many of which are addressed by using something like CoffeeScript.
It doesn't really matter, browser games are the future and no matter how complicated it's going to be, somebody will be good enough to write them. We already passed "demo era" and some games are really good, like https://chrome.google.com/webstore/detail/bhoaojooagiaaiidln...