Comment by Sophistifunk

We spent 40 years chasing down and eliminating every redundancy, buffer, backup, and alternative in the name of "efficiency" all the while globalising every system we could. Even the slightest hiccup is liable to bring such a fragile system down, and this was not a slight hiccup. And as supplies dwindle throughout the system, demand increases at every interface point, not just the end-users. This machine is not going to come back up as it is. It just can't. We need to start putting those backups and redundancies back in place, which will cost throughput, latency, and (fuckloads of) capital.