This will call json_variant_sensitive() internally while parsing for each allocated sub-variant. This is better than calling it a posteriori at the end, because partially parsed variants will always be properly erased from memory this way.
This might make things marginially faster. I didn't benchmark though.