The original model of querying, doc by doc, whether doc ID exists to then insert/update the corresponding doc does not scale