Not sure. Probably networking/distributed systems textbooks or network/telco hardware docs prior to 2000-2003. That was the conventional wisdom back then. Dig around in IEEE and ACM articles of the period and you will find lot of commentary about the broadcast/scaling issues.
Back then Data Centers didn't exist as they do today. Most large system architects didn't believe emerging new startups like Twitter/Facebook/YouTube etc would survive because the scaling problem, that many to many broadcast required, hadn't been cracked yet.
There was only email and IM back then and no one was broadcasting to millions of people every 2 mins. Because it would just crash everything.
That's why large companies at the time like Yahoo, Google and Microsoft etc totally missed the social media boat. They didn't believe it would work and were already running into scaling issues with their own systems.
It mostly all luck that it worked out. Telcos worldwide upgraded pipes and Moore's law drastically dropped cost of processing exactly around that period making data centers/cloud possible. And all of a sudden we are in a new world were many to many mass broadcasting is a reality.
Beyond the CS side to this recommended reading would be Howard Simon's concept of Bounded Rationality, it's implications, work arounds etc. It's based on studying large orgs and how they fail when human cognitive limitations get hit.
That kind of approach probably needs to be scaled up/developed to handle the broadcast issue.
With everyone broadcasting it's easy to hit mental bandwidth limits. And then mistakes start happening. People react to each others mistakes, hide them etc. Fixing thing get more and more complicated and then you have a runaway cascade. It becomes a trap.