I'm no coatings chemist, but varnish is an organic material and I practice biochemistry every day. I also run a high-resolution microscopy laboratory, so I spend a great deal of time on electron microscopes. What I see there likely applies here.
The prevailing wisdom is that sanding creates scratches, and therefore creates much greater surface area and a more complex microarchitecture of the surface to which the next coat of finish will be applied. Even so, the dimensions of this microarchitecture are much greater than the molecular dimensions of the finish. Therefore, it seems unlikely that scratches alone are responsible for adhesion. In other words, adhesion of one layer to another is not just mechanical. The solvents in a finish that is compatible with a previous finish (such as one coat of varnish over another) can dissolve a bit of the surface of the previous coat, thereby allowing the two layers to intermingle and cure together. "Hot coating" works because the previous layer of finish is incompletely cured. Because total cure time is long- much longer than the time-to-next-coat listed on the can- hot coating should be doable for a long time, with increasing time after the last coat meaning less and less ability of the new coat to merge with the surface of the old one. Thus, I believe that sanding scratches, by creating this complex microarchitecture, do two things: (1) scratches increase the amount of surface available for a bond between the two coats, and (2) they open up the previous layers of finish that may be less fully cured than the surface (at least within the first year or so). Increasing bond area by increasing the complexity of the surface should become increasingly important over time after the previous coat is applied.
This is not to say that mechanical adhesion is unimportant. Finishes that polymerize as opposed to simply drying will eventually fully polymerize and subsequent layers can no longer solubilize even a small amount of surface. In this case, the microarhitecture of sanding scratches should prmote mechanical adhesion.
There's one more thing that applies particularly to older finishes. Surfaces, especially as they get older, become increasingly contaminated, and some of that contamination is not just sitting on the surface, but rather chemically bonded to the surface. Sanding cleans that old surface (well, removes it), removing impurities that may be well bonded with the old surface.
Bottom line? Yes, hot coating works (I usually do exactly what Andre describes), and "less-hot" coating- longer time between coats- should work as well, but it should become less effective as a finish gets progressively older. But as this happens, sanding still produces a more complex surface for mechanical adhesion of one coat to another. So as a finish gets older, sanding becomes increasingly important for promoting chemical bonding if still possible, and mechanical bonding in any case.
Michael