> Finding ~50 dots arranged in a (very loosely defined) circle, from any projection, of a dense set of 2 trillion of them is very plausible.

It depends on the size of the circle, though. The smaller the size, the more likely the probability is. But that’s only for a particular combination of 50 dots. Now we have to average out of all possible circle sizes and all combinations of 50 dots. Can someone do the math (or the simulation)?