We noted some of the downsides of using WebAssembly in this blog post 2 years ago.
"50 times faster" is also a massive exaggeration - from what I've seen it would be more like 20-50% faster in certain specific cases. It also would likely make no difference at all in many cases, since a lot of games aren't CPU-bottlenecked in the first place. Then it would be a huge amount of work to rewrite the entire engine in WebAssembly, with high on-going maintenance that would permanently slow down development and therefore reduce the number of new features we could work on. Given we already made some huge performance improvements in the C3 runtime, all just using JavaScript, I don't think it's necessary or worth it.
We do already use WebAssembly in a few places though - notably the physics engine uses a WebAssembly build of Box2D, so the physics engine is already native-speed, Advanced Random uses a WebAssembly module for noise generation, and the WebM Opus encoder/decoder is based on WebAssembly.