You presume to know more about the implimentation details than I do. I expect there is not a "list" of mutations. Rather, an algorithmic mechanism for mutations which, unsurprisingly, employs somewhere at it's heart, a random number generator to decide which direction to tweak a mutation, not from a list, but from a primative understanding of what it means, in detail, to mutate.
There have been experiments that have proceeded as you are imagining, but they weren't nearly as interesting. The presumption that you know what all the acceptable mutations might be ahead of time, being the simplifying assumption that makes such a list possible to contemplate.
I am trying to understand the program based on the article. I have no knowledge of the program beyond what is stated there. As I understand the article, simpler functions mutate into more complex functions in the program. The article mentions these "simple functions" in several places without going into detail. I would think the programmers had to program some functionality into those simpler functions and also create a routine that generates the mutation or allows the mutation to occur (yielding the "random mutations"). That's what I'm trying to understand.