The problem is that in the end, you won't see consensus on a secondary language from multiple browser vendors, in the end something that breaks down into JS/asm.js is probably more practical.

To me, getting webworkers to a better position of usability is a far better goal. As it stands, you have to write your worker completely detached from your main script/code... separate build target, separate calls, different model from node's IPC which reduces reusability. This is mainly in that I feel a bigger issue with JS performance for critical tasks in the browser will be alleviated by better channels with web-workers.