In popular literature, one of the more under appreciated factors in various models is the moisture carrying capacity of the atmosphere. Basically as the mean atmospheric temperature rises, the more moisture it can carry. This leads to the formation of clouds as different temperature air masses meet, and precipitation when the temperature drops below the dew point.

When you run different climate models, if the clouds form in the stratosphere the overall temperature goes up (more CO2 below the clouds to reflect IR back to the ground) if the clouds form in the troposphere the temperature goes down (clouds create a higher albedo and more energy is reflected into space rather than absorbed).

One of the papers discussing inter-glacial periods expressed essentially the same view as the Harper's article which was that kicking off an ice age would require a tremendous amount of material in the troposphere to reflect sunlight, and a bunch of water to condense into snow to boost the albedo of the surface to prevent it from absorbing solar energy and re-radiating it as heat.

In the 80's, when people looked at the possible after effects of a nuclear exchange between the US and the then USSR, it was suggested that the level of soot and ash thrown into the air would cool the planet to a point to enshroud it into a "nuclear winter." Later papers have looked not at nuclear exchanges but simply volcanic activity or firestorms of one form or another.

The duration of these effects is tied to both the amount of sunlight obstructed and reflected and the resulting weather effects.

Using those models it seems plausible that prior to the presence of humans, if the global temperature rose, it would dry out vast tracks of forests and grasslands in the mid-lattitudes, those tracks would be set on fire by lightning creating a tremendous soot load on the atmosphere. Once the soot layer forms, the temperature falls, and the atmosphere drops its moisture as precipitation. If the temperature is low enough it becomes snow, and gets compacted into ice. The ice become glaciers and it takes a while for the cycle to reset.

For me, it is an interesting aspect of climate dynamics on the water cycle as a feedback loop sort of problem. I have yet to see any paper that suggests the 'venus mode' (runaway greenhouse effect leading to boiling the oceans) is any more or less likely than the 'snowball mode', other than we have a geologic record of the planet going snowball mode on its own several times over the last million years.