Any one error only affects a portion of the network. For example a problem might take Australia or the American East Coast offline, but that doesn't significantly affect the rest of the network (unless they try to reach affected regions). That was the design goal and it works perfectly.
What you want is entirely different. The European power network for example is designed for (n+1) redundancy: any one equipment failure doesn't have significant effects. If you take that a bit further you could include misconfigurations or even allow entire companies to fall out of the network. But each level of assurance requires more overprovisioning to compensate for failed equipment or lost capacity. And overprovisioning is expensive.
What you want is entirely different. The European power network for example is designed for (n+1) redundancy: any one equipment failure doesn't have significant effects. If you take that a bit further you could include misconfigurations or even allow entire companies to fall out of the network. But each level of assurance requires more overprovisioning to compensate for failed equipment or lost capacity. And overprovisioning is expensive.