If you hold a measured loft constant, but you add more fill to it, density goes up and insulation goes up. That is misunderstood as higher density is warmer, which is not true. If we reverse it and hold the fill quantity constant but reduce loft, density goes up and insulation goes down. The correct conclusion would be that on a per quantity basis, higher density is less warm.
These are kinda the rules that tell the full story. If I fill an open top box with down fill and let it loft to its maximum potential, that is the absolute maximum insulation that amount can achieve. If I compress it within that box, density increases and R-value decreases. If I close the top of that box but keep stuffing more down in, density and R-value will increase. If I then open the top and let that amount loft to full potential, R-value will increase even more, with the increased loft, while density goes back to zero.
If we’re viewing this in a vacuum, an amount of fill has the most insulation for the least weight at its lowest density. If we take everything into account, a certain amount of density is required to maintain coverage and make up for compromising circumstances, but that density always comes at a cost to warmth to weight efficiency. Ideally, density should be catered to the use case so that there is enough to maintain coverage, but not so much that warmth to weight is reduced too far.
If density is very low, in the real world, the fill will shift around and leave gaps. These gaps will introduce convection, which will remove more heat than is gained by the increased warmth to weight efficiency of low density. If density is very high, the amount of insulation you get from an amount of fill is going to drop. You’ll have a large buffer for getting it wet or putting it into a lot of compression, but you just won’t get a lot of insulation from that amount of fill.