It's not academic writing, just a short explanation on why something that was expected to be a major advantage of QUIC is being removed from the spec for now, due to having a negative impact on Youtube. (Of course that shows that they're able to revert bad decisions, which is definitely an advantage over TCP...)
Though there is an interesting rebuttal in the related mailing list thread [1] suggesting that the feature is sound as-is, but was being applied in the wrong way in these experiments. Specifically, that the largest expected gain is for the client-sent packets very early during the connection setup.
[1] https://groups.google.com/a/chromium.org/forum/#!topic/proto...