Ah I must have skipped by that bit. And presumably the memory-hard algorithm is CPU-easy enough that this won't be a problem. That leaves me wondering "why not just use a JS memory-hard routine in the first place, but maybe I've shouted over that too. I'll give TFA a reread later.
I think it is still compute-heavy if not compute-hard, and stuff like WASM SIMD makes it less compute heavy (not really available in JS). And also, for memory bound problems, you probably don't want the JS GC needing to clean up afterwards when WASM gives you that control.
I assume that SIMD acceleration is usable with WebCrypto, but I have no evidence either way. It's also not part of WebAssembly so there's that to consider too lol