His point, as suggested by the comments agreeing with him and the claims that it's 'obvious', is that the parameters could be small. The problem with that is the parameters look large. Adopting a distribution and calculating it out gives much broader predictive distributions. You can have all the parameters not be that small but small enough, combined with minimal uncertainty, that on many draws from the posteriors, you get underpopulated universes once all the uncertainty and probabilities are integrated. What he wants to take by theft from individual parameters, handwaving them away as being small and thus causing the paradox, can be obtained without any trickery for a wide range of parameter values and 'margin of error's simply by the nature of the pipeline process producing a wide range of final outputs. It's akin to confusing the confidence interval around a mean for the predictive interval.
I asked how his statement could be true while your statement could be false (which would show that even if his statement was granted then your statement could be new and non-obvious). You replied by claiming that his statement isn't justified while your statement is. This is a formalization of an intuitive argument, not a surprising dissolution of a paradox.
In other words, if other researchers had considered his statement and dismissed it without doing the sort of analysis being formally written up in this paper, then those dismissive researchers would have been making a mistake.
As I said elsewhere, I think these sorts of formalizations are really very helpful. Perhaps, given the sloppy thinking in this area, even more helpful than novel results. I generally think review articles are distillation is vastly undersupplied
