In worst-case scenarios, I think the Robinson Crusoe approach is futile. You can't store enough beans, enough bullets. Networks will keep us going, and success will depend on how robust your network is.
For the nearest real-world example of the calamaties you describe, look to Argentina in 2000. Very quickly, elaborate barter networks emerged, and people kept going.
The cake is always in the oven. The trouble is we never know enough to accurately predict anything. I set no more stock with the gloomy predictions than with the good. Things are very volatile now, and because of strong interactions between perception and reality, the prediction itself be variable in the model on which the prediction is based. Not easy.
Market Data 
Markets 
AI
