This is obviously not true for every case, but most of the time designing you’re working on systems and refining them. So when you see one case where there’s a problem, it’s easy to fix that by tweaking the system. But then there’s a big chance it would ripple out and create new problems elsewhere. See design should not only be judged by the problems it solves, but by the problems it creates. It’s often useful to try and see if the root of the problem is actually related to a system. If it can be fixed by changing the content (e.g. level design), it’s usually an easier and safer fix. Especially later in development.
For Circles, there were many cases where a specific hitch in a level created the problem. It was easy to see a solution to the system to counter-act it, but it turned out to be way more effective to change the given level and prevent the problem altogether.