Usenet was a set of fiefdoms mostly administered by academics in CompSci departments, and proved utterly unequal to its first real crisis*. Distributed systems work great as long as they're new and everyone is participating in good faith most of the time. In adversarial situations, they're rarely able to adapt flexibly enough, partly because the networked structure imposes a severe decision-time penalty on consensus formation. A negligent or malicious attacker just has to overwhelm nodes with high betweenness centrality and the whole network fails.
Immediately following crises everyone talks about making the network more resilient and so on, but it never fully recovers because everyone intuitively knows that establishing consus is slow and bumpy, and that major restructuring/retooling efforts are way easier to accomplish unilaterally. So people start drifting away because unless there's a quick technical fix that can be deployed within a month or two, It's Over. Distributed systems always lose against coherent attackers with more than a threshold level of resources because the latter has a much tighter OODA loop.
Immediately following crises everyone talks about making the network more resilient and so on, but it never fully recovers because everyone intuitively knows that establishing consus is slow and bumpy, and that major restructuring/retooling efforts are way easier to accomplish unilaterally. So people start drifting away because unless there's a quick technical fix that can be deployed within a month or two, It's Over. Distributed systems always lose against coherent attackers with more than a threshold level of resources because the latter has a much tighter OODA loop.
* https://en.wikipedia.org/wiki/Laurence_Canter_and_Martha_Sie...