In my experience, most of the best philosophers working on the philosophy of religion are in fact theists, due to fairly obvious selection effects. Of course, most philosophers of religion, good or not, are theists, but I think in the upper echelons of the discipline the disparity is even more stark. Atheists who get into the field tend not to be very good philosophers, with a few honorable exceptions.

So a disproportionate number of the most careful and clever arguments (which, I think, is what David meant by "best arguments") in that field support the theistic side, the wrong side. If the only thing I knew about an article on the philosophy of religion is that it is extremely well argued, I would consider that evidence that its conclusion is false. Does this violate conservation of expected evidence?

Note that there's a distinction between all the arguments that exist in some Platonic sense and all the arguments that exist in published form, and that there's a distinction between arguments that are "good" in the sense of significantly raising the probability of their conclusions and arguments that are "good" in the sense of being clever and carefully constructed. In both cases, I think David was talking about the second option. He was attributing to Yvain the claim that most clever and carefully constructed arguments out of the set of published arguments are for bad ideas. I don't agree with this claim but I don't think it violates basic rules of probability.

Unless I misunderstand, only if he hasn't already updated on how many believe in the idea.