Stability follows from N′(T)=−A+2BT<0 at the lower root. They merge at T=A/(2B)≃8.97K and Fmax=A2/(4B)≃5.47Wm−2; at equality the equilibrium is marginal, not linearly stable.
Assuming the usual logarithmic forcing extension F=3.7log(C/280)/log2 gives a stable branch only for C<Ccrit≃7.8×102ppm. This “maximum” is the supremum of stable concentrations in the stated quadratic, linearized model. It is not a physical prediction of a real-world runaway threshold; extrapolating its feedbacklaws far beyond their small-change regime is an additional assumption.