Yes indeed, this was my initial idea. But the main drawback is that it would need to work in bulk mode as well as on a per document basis. It's probably feasible though, I'll have to look in the code how I could use the same bulk format that the _bulk endpoint accepts.