I guess my presumption was that algorithmic trading was a very tight feedback loop, with as many controlled variables (i.e., GC) as possible, so I think it just subverted some of my misplaced expectations.