If I had to relate this to an HN related topic, I'd say this is closest to the theory of "technical debt". The further your code base moves away from the origin, the more difficult it is to address issues in the foundational code. Fudging around with the basic building blocks of life is extremely risky in an evolutionary perspective, gains can be made but they are going to be smaller (less advantageous) than gains possible by fudging around with top level code. By making base level changes, you effectively halt top level change until everything is proved out. By the time a base level change would have stabilized and the gains realized, the more nimble organisms making top level changes will have long since outcompeted the base changer into oblivion.