Friction is extraordinarily complicated and not well understood, especially not by me.
There seem to be two mechanisms at work, as I understand it. One is phonon exchange -- when you have two surfaces (or a gas and a surface) moving past each other, the atoms in each collide and in effect pluck at each other and set each other vibrating. These vibrations -- called phonons -- absorb kinetic energy from the moving atoms and are apparent as heat (which is, very crudely, atoms vibrating).
The other mechanism is the interaction between shells of electrons in the atoms of the surfaces/gasses. These have minute fluctuations in their density, and if you get a denser area in one shell it will repel the cloud in an adjacent atom's shell. This produces an area of positive charge, which in turn is attractive to the negative charge in the shell that caused the fluctuation in the first place (I think that's right -- seems a bit topsy to me, but that's electrons for you). These are called van der Waals bonds, and if you get atoms moving past each other the bonds once again tend to pull the atoms together and kinetic energy is removed from the motion of the system as they pull apart.
If you increase the rate of these inter-atomic interactions -- either by increasing the speed or increasing the density of the materials involved -- then the energy transfer will increase, friction will increase and heat will increase.
Perhaps. I don't understand how atoms 'collide'-- my knowledge of what actually happens at a quantum level is not good -- or whether this collision involves anything other than the van der Waals bonds. I do think that the electron model is attracting most interest at the moment, as it creates the possibility of surfaces where you can vary the coefficient of friction by changing an electric current flowing through them.
R