Some, a bit more detailed, thoughts.
"The CLIMBER-2 intermediate model includes a simplified atmosphere model which allows for runs of long periods.", or something like that.
The model used in this study is light compared to the full scale models that may lack some feedbacks. They're better suited for studies involving shorter periods ~ the 160-250 (maybe a thousand) year scale we have plenty of data to verify them against. And as they have been developed from meteorological models, they might use inordinate amount of resources calculating the atmosphere... Thus slow glacier feedbacks, such as have been postulated to be of influence f.e. in the Green Sahara periods, may not appear at all.
As we have seen marine glaciers may disintegrate very rapidly in Antarctica, it's prudent to test the various developing glacier models in these smaller climate models. The model used in this study should probably have a dynamic ice sheet model for Antarctica too, but they've opted to think the ANtarctic Circumpolar Current is a sufficient barrier. And the regolith-model used for northern glaciers doesn't apply in the southern pole, there are no such soils present under the evercold Antarctic icesheets.
Like *sidd*, I'd be interested on the more detailed results for Eemian-present. Could their glacier growth-decrese model send enough icebergs sailing shouth to produce the Green Sahara and glacial temperature variations observed from proxies?
They apparently get the large scale structure of atmosphere being correctly formed, with monsoons and other general fronts, so maybe the model would show a persistent atmospheric river or an extra monsoonal effect over in dry Africa. The 10-16 levels mentioned in the model description is enough to resolve these then. Didn't quickly find the specs on the ocean model used, but it's of GFDL make and also used by them so it can't be extra bad. I don't know enough of the other modules so no comnent on them, let the results speak.
The high spatial resolution on the atmosphere in more complex models can be used in f.e. hurricane studies, EMICs (Earth model of intermediate complexity, if correctly remembered) do not find these (or at least didn't some 5? years back), and I'm of the opinion that ocean models should have (containing large amounts of latent heat) should have almost as many layers as the atmospheric modules, but this would likely mean lowering the amount of layers of atmosphere, to conserve disk space and computing time...
Ok, I think that's enough of a ramble on this. The study is such it presents imho a challenge to the teams working on the massive models. Had to read the realclimate write up first to get into the article, not finished with it yet, maybe in future.