Thats what the analysis at nanex points to. It's basically denial of accurate timely quotes across exchanges. Don't know if was an actual "attack" and weakness in the system or just a random event and weakness in the system.
Applying my electrical engineer perspective to the Part 4 graphs, it looks like this scenario may have happened:
1. Somebody's trading software was issuing quotes based on recent quotes, forming a feedback loop.
2. The feedback loop was calibrated in a half-assed empirical way by somebody who did not understand control system theory, and did not even know they did not understand it.
3. The overall system developed a delay somewhere in that feedback loop.
4. The delay exceeded the phase margin of the system, turning their half-assed negative feedback into positive feedback.
5. Positive feedback -> oscillation. This triggered massive quote volume, keeping the backlog high and therefore keeping the oscillation going.
6. The quote volume put loop delay on quotes on other issues, pushing them into negative phase margin, causing the trading software to oscillate with them too.
7. Avalanche!
The solution is to add a 50 picofarad capacitor between the traders and the stock exchange. ;-)
http://www.nanex.net/20100506/FlashCrashAnalysis_Part4-1.htm...