Surprising Adobe killed it *that* fast - RIP.
As for the current discussion, I believe WebGL is the right solution. Yes, there are security concerns, but some of them would have existed with *any* GPU-accelerated solution with remote access, including any current handheld application that wasn't checked properly. And I don't think there is anything inherently unfixable about any of WebGL's problems although it may pose a variety of vulnerabilities in the short & medium terms.
Yes, you can do something decent-ish with CPU-based solutions today. It's also fundamentally unscalable. Yes, Flash's GPU acceleration might have been a viable solution, but Flash also has security (and privacy) vulnerabilities of its own and I doubt GPU acceleration didn't add a few. And yes, HTML5 implementations vary too much, but thankfully WebGL only uses a tiny bit of HTML5 functionality and a lot of JavaScript, which should be much more compatible.
Don't get me wrong, I don't think WebGL is a perfect solution, but I do feel that if implemented properly (which I'm confident it will be eventually), it's good enough. Just like I don't like JavaScript is very good but I think it's good enough for web computing - if you need more than WebGL or JavaScript, you can always leave the web and write a native application.