Actually, don't get me wrong. I think HFT is great (and I thought your HFT Apologist series was excellent). It reduces spreads greatly. There is nothing intrinsically wrong with prop firms, market makers, etc. But it is also an increasingly complex system and risk is part of the business. My meta-point (certainly not particular to Knight) is I don't think we understand all the consequences or the risks yet stemming from the fairly saturated and very competitive business, and the continuing arms race. There is also that unpredictable and fallible human component and how it reacts or affects the automated agents. There was even a paper on how as we approach zero, there may be brand new "relativistic" arbitrage opportunities: http://www.alexwg.org/publications/PhysRevE_82-056104.pdf. The Knight incident seems to me a reminder that there is much more to be done in risk management, more robust modeling, and defensive software development technologies in fintech. I don't know when the incentives will be there to invest in such things.

Yes, exactly, but it is still risk. They certainly haven't lost their money yet. If everything works out right, they shouldn't have to. But the financial world is much too complicated to say that there is minimal risk for any party, even if everything is in Treasury bonds.

Moreover, I would suggest you read Johnson et al's research on mini-flash crashes (http://arxiv.org/pdf/1202.1448.pdf) if you hold that the markets are stable dynamical systems.