I look at it as building software on top of an extremely complex system. It's just not going to move as fast as anybody likes it to, and often the best course of action after considering a new feature will be to leave it alone because it's not business critical and worth the risk.