Supporting WASM feels like a moving target. We get support running for one version of Emscripten, then it stops working in the next, or we have difficulty getting the same environment running on a different host system platform.
I have not seen any performance benefits to WASM in my testing (though admittedly, these tests were more GPU- than CPU-bound) but with much larger binaries, more difficult debugging, much longer compile-times and the instability in maintaining support for the toolchain, it was dropped from haxelib releases.
However, I do think that it may be a valuable target in the future, and have no need to break the source compatibility we have so far. I believe our JS HTML5 target is better all-around right now, but some day in the future, WASM may surprise me and it will be stable, fast and the best way to go