[jamespitts]

During the 2008-9 crisis I learned that the failures and responses could be understood in the context of mass computational problems.

Before a crisis, the system moves along w/o having to fully process everything; this is because a low-fidelity understanding of value and status of securities and agreements is sufficient for participants to operate. However once the crisis has begun, new dire weaknesses and new structures require much higher fidelity to guarantee safety. The participants find that there is much more to compute than can be computed in any reasonable timeframe.

This is why the monster-sized purchasing programs by lenders of last resort are somewhat effective; the are enabling everyone to function by delaying the difficult financial computations.

[esoterica]

The problem was uncertainty about the inputs, not insufficient computational power to crunch those inputs. If computational power were the bottleneck there would have been a mass spike in server purchases and a run on compute capacity, but clearly there wan't.