I will be honest with you: I don't know the technical details of this. I have a only hazy idea of the basic problem: It has something to do with the way the algos are entangled with each other. They are too intimitely entangled now to make such "tweaking" as you mentioned possible. Some of it has to do with the derivatives market: They are all programmed to react on a microsecond-order level with each other, and to react to each other's reactions, and to react to the reactions to the reactions, and so on. It has passed beyond a level of complexity that the human mind can even comprehend, much less unravel.

Another aspect of the problem is all the most advanced computers and systems now use souped up "neural network" tech. Basically, the connectivity is modeled on a human brain, and the systems "learn" automaticaly and self-modify the way a growing human brain does. We can't "re-program" their behavior because they aren't acting on simple programs; the are self-iterative and self-modifying, and they have spun off to a level of complexity and speed that is, again, beyond the realm of their creators to handle.