Yeah, the only thing we can really learn from this paper is that the authors aren't very familiar with protobuf. Anybody who claims it takes 10 microseconds to decode a 1KiB protocol message is either intentionally constructing the worst-case message or doesn't know what they are talking about.