“…Let us turn now to the higher resolution information contained in our results, beyond the limit of N −1 F N . In the subcritical regime, θ ∈ (0, θ c ), we show that F N and log Z(1, 1; N, N ) behave the same down to order N 1/3 , and thus conclude our theorem by known asymptotic results for the later random variable [BCR13, KQ18,BCD20]. In the supercritical regime, θ > θ c one might hope that the maximizing points will be order one apart, and hence the log(N ) behavior for F N makes sense from standard extreme value theory.…”