Spherical cows are an in-joke among physicists.
The idea is, if you want to understand the physical world and make predictions about it, sometimes you must simplify your assumptions.
Taking a very silly example, if we wanted to calculate the speed of a skydiving cow, we might assume the cow is perfectly round and then make calculations about the forces acting on it. Sure, we won’t get a perfect picture of how an exactly ‘cow-shaped’ object acts but the model is good enough for us to make basic predictions. ‘Spherical cows’ encapsulate this way of simplifying problems so we can understand them more easily.
I love this sort of simplicity and enjoy bringing some of the same thinking to my work on evaluation. Theories of change are one example of how we sometimes take something big and complex and condense it down to its constituent parts.