But wait...it's a client speed up optimization. All in all it would cause the same amount of traffic just reordered to start pages render sooner.
As for Joe Human, his reaction time is at about 20ms. Hence @5% improvement would be theoretically "noticeable" in a 2 seconds page load. But unnoticeable if Joe Human would have to observe it relative to 2 seconds total. Likely even with a stop watch Joe H. would be in "error area". And 100ms would be an improvement on a 20s load which would challenge patience of any Joe H.
From improvement description:
"The primary drivers are: preloading images sooner, more aggressive use of idle network time, dynamically changing resource priorities, deprioritization of preloaded resources, and reduced bandwidth contention among images."
From TFA:
"when you add up those saved seconds across all Chrome users, it totals to more than 510 years of people’s time saved every week."
Now that's convincing argument.
I admit that squeezing that 5% of already well optimized code is a great achievement (technically) that may be head-deep in the law of diminishing returns (practically). But G$$gle can afford it.