But most of the surface area of the HX is cooler, where the gases of the fire side are already literally thousands of degrees cooler than where they entered the HX, but the delta-T across the HX to the water side is now much lower, and that temperature is governed most-predominantly by the return water temp. To the extent that the return water to output water delta-T is bounded, the amount of HX surface area that is below the dew point of the exhaust gases rises and falls with output temp, but the amount of condensation derived is still governed by primarily the return water temp. The size of the flame, and the output temp are only second-order effects.
Below some firing level there isn't sufficient turbulence on the fire-side for good heat exchange in the low fire/water delta-T sections. Laminar flows create an insulating boundary layer that limits the amount of fire-side gas contacting the HX thus reducing the condensation. This is the technical reason why it's difficult to build a condensing boiler with more than about a 4:1 turndown ratio. At any given return water temp in the condensing zone the sweet-spot is usually somewhere in the lower 1/3 of the flame modulation, but not necessarily at the lowest. Cranking the flame lower than min-mod on most of them they run into laminar flow issues and efficiency falls off very rapidly with fire size as more of the gas volume then escapes without coming into direct contact with the cool end of the HX.
An oversized boiler would be more likely to short cycle, but would deliver the heat at a lower flame modulation, with ever so slightly higher condensation for the same volume of exhuast gas BTUs. At any given return water temp it would still deliver about as much condensation as an appropriately sized boiler, but would be giving up more in fixed losses with every flue-purge & ignition cycles, delivering somewhat lower overall efficiency.