This was an experiment to both 1) learn how market making works and 2) figure out what is so hard about it. My main insight from this process is that order management--making sure only the orders you want are placed, and fast, is a pretty complex software problem. The algorithm itself seems to be one of the easier things to implement.
I find that the best way to solve that is to make the strategy define its desired state, and then you have another component that diffs the current and desired states to identify the necessary transactions to send out as well as their priority based on the nature of the diff, then finally go to a scheduler.