JIT supply chains have always been extraordinarily fragile.<p>We saw a small version of this global situation with Katrina when supply chains emptied out in the South. Typically there is only 3 days worth of inventory in most grocery stores of either perishable or non-perishable just because of consumption rates out-the-door by consumers. This was hidden when everything upstream "simply worked".<p>The problem is that each node, by JIT norms, tries to keep ZERO inventory or close to it. So the standard model is that as inventories drop in retail, which is why scanners were introduced replacing stuck-on price tags that were the norm in my youth, and hit a defied threshold, an order is placed electronically to the supplier for that item (or items). Then if there is any inventory, it gets shipped but again the ideal is ZERO INVENTORY, so that order from downstream triggers an order to upstream. All the while there is "in-transit inventory" that will arrive first.<p>The inventory systems used are statistically/probabilistic so they "predict" futures based on past consumption and trigger orders. It's not unlike a PID controller but not that sophisticated. The UNIVERSAL PROBLEM with all such systems (and that includes option risk/pricing including all derivatives) is that the assumption made is that the statistics are CENTRAL and thus returning to a mean, and that implicitly REQUIRES that each element/transaction/etc. MUST BE INDEPENDENT OF ALL OTHERS - otherwise the assumption of centrality is not mathematically valid. And we happen to know that things like markets are not independent in this way - that's the implication of Mandelbrot finding that markets are fractal - the directly implies long tails and non-independence.<p>So if anything creates a situation of major cross-correlation by something like: market consolidation, market juicing with QE and then with a cherry on top of a universally affecting pandemic that suddenly makes everything codependent with one variable: the pandemic and its cause world-wide, then all bets are off for inventory models and risk modeling like Black-Scholes. Monopolies also create the same type of universal cross-correlation/co-dependency that breaks every type of statistical/probabilistic model previously used.<p>I've been expecting this to happen eventually. Sadly I've never come up with much of a defense against it or the political/totalitarianism likely to follow it. At least no a universal solution everyone could use.