It's important to understand that many manufacturing industries operate on a "JIT (just-in-time) manufacturing" model[0], where goods used in a manufacturing process are only ordered and received as needed, minimizing inventory and storage costs. (e.g., instead of having a warehouse full of screws sitting around waiting to go into products, you order X screws from another manufacturer every month to fulfill Y number of effective orders) This means that many industries (especially the automotive industry) operate on a constant stream of supplied products, in many interdependent chains.
There are two downsides to this approach:
1. The JIT model has to be at least partially predictive. You can't necessarily order components right as you receive orders, because it takes time to produce those components (remember, no one has warehouses full of those components ready to ship, because it's often JIT all the way down) and you don't want to keep customers waiting as components get shipped to you. Instead, you need to at least somewhat predict how many orders you expect to fulfill in order to balance between keeping customers waiting ("shortage") and having excess stock laying around which might never sell (waste)
2. The other main downside is that JIT production is highly interdependent on other industries and can be fragile. If you expect to be able to order X screws this month from your supplier, but some external event causes them to only be able to supply you with Y screws (Y << X), you're shit out of luck. You literally cannot produce your product because the underlying components simply do not exist, which leads to a shortage of your product
Both of these downsides come to a head in the presence of some global event that causes uncertainty and inefficiency. If a low-level JIT process depended upon by many industries suffers from an inefficiency (e.g. workers can't come in because they're sick), many higher-level processes start to slow down (e.g., no one has screws to build their products with). This effect is multiplicative across industries, since many manufacturers specialize in specific products and downtime can affect many "downstream" clients.
Along with this, such global events can dramatically change customer demands, and when you've only predicted needing to produce X products but you now have Y demand (Y >> X), customers are out of luck because the products literally do not exist to be sold (shortage).
Considering that computer chips are now present in almost all major consumer products these days, it doesn't take much uncertainty and inefficiency at the "base" for a domino effect to significantly slow down a lot of industries.