I am currently fascinated by the process by which one selects a hypothesis to test over other hypotheses, and that one can't test a hypothesis all unto itself. Funnily enough, even that is a hypothesis.
We have some concept of the universe we want to test -- a hypothesis -- and we select to test it out of several others. It all seems to match up after observations are made, but that matching may only be us looking for positive reinforcement of our own idea. So you also test a second hypothesis at the same time, the Null Hypothesis. The Null Hypothesis states what evidence would prove our initial hypothesis conclusively false.
But still, in the midst of this, there isn't a step by step process by which we choose a hypothesis -- there is no mechanism, no real way of knowing how to choose the best hypothesis. There are guidelines, but ultimately, science doesn't care how one chooses an idea to test. All science really is is a method for testing the "soundness" of an idea.
And even when the idea is validated, we often later will recount, reform, and rephrase our understanding of the universe. And... well, that fascinates me. It drives the point home that science is, while a rational process, is also an arational process at its heart. And it makes me wonder: Are all bodies of knowledge similarly arational? Euclid didn't have a method for choosing his postulates. Aristotle didn't have a method for distinguishing between his "Causes" -- it was essentially just really smart people pulling stuff out of their ass. If not math, science, or philosophy, what is fully rational? Logic?