Not all down is equal. Down is a blend between the fluffy down clusters and feathers.
The fluffy down custers are composed of very fine substructures or microstructure. You can't compress a poluyester microscope slide, nor a glass microscope slide — because they break. You can compress a mass of polyester fibers or a mass of fiberglass fibers can be readily compressed. This is a reflection of thinness. If you have had the class in "Strength of Materials", thinness in materials promotes easy compression (actually, bending and/or torsion and/or buckling) of a mass. For example, a steel fiber mass, such as a scouring pad, compresses easily, but a solid one piece steel bar doesn't.
Feathers are composed of microstructural elements that are generally thicker than the microstructural elements of a down cluster. Not only are masses of feathers harder to compress because of this geometric factor, the microstructural elements of feathers, particularly the thicker and stiffer stems, will break upon compression.
It is the very microstructural fineness of down microstructure that is a basic building block for compression without damaage.
There are other factors in addition to the fineness. Humidity is often a plasticizer for organic structures. Heated steam, for example, will let you permanently bend wood into fine furniture or sculptural or tool items. Without the plasticizer effects of steam and heat, you can't generally do that with wood creative permanent bendings of large amount.
So there is a physical basis to the cautions of humidity and compression for a long time. You would expect that under humid conditions, the down clusters when highly compressed could more or less permanently assume a less fluffy configuration related to the plazticizing activity of humidity (water vapor or cold steam).
This humidity effect is postulated to be possible on general principles known to apply to other organic materials. I have no direct experience of humidity effects on the long term compression of down.
There was a masters dissertation on the mechanics of down written at the Georgia Institute of Technology within the last 5 years. It focused on the elastic compression characteristics and models for use in analysis of down deformation.