Can anyone clarify the CO2 band saturation thing?

CO2 increases broadening because it is heavier and there is more momentum exchanged in CO2-CO2 collisions than in CO2-N2 (or CO2-O2) collisions.  The difference in mass is about 30%, hence the 30% increase in broadening.  See here.  Other trace gases will only increase the broadening if they are heavier than N2/O2, but there aren't any of those at significant concentration for this to be a factor. 

I forgot the first parts, sorry. 

The cold parts they are taking about are from about 6km up to 10 km, really the top part of the troposphere.  This is why deep convection is so important in the global heat budget, since it lifts heat above the layer of the atmosphere where most of the IR absorption occurs. 

It's a collision thing, where the impact between molecules gets molecules into excited rotational levels.  These rotational levels couple into the vibration spectra, broadening the lines.  Heavier molecules colliding lead to higher rotational levels, and the higher the level (and more rotational modes that are active), the more options the molecule has to relax into once a vibrational transition occurs.  Since rotational level changes take far less energy than the changing vibrational energy levels, and there are so many rotational levels available compared to vibration, you don't see large shifts in the wavelength of the vibration transition.  Instead, what you see is the wavelength of the vibrational transition gets smeared over a larger range.  But what is really going on is that you are seeing the effect of more rotational levels being active in the process.

Sort of.  A real spectroscopist would beat me to death with his Jerzy-Turner monochromator for explaining it like that. 

Pretty much any good p-chem book will have a discussion of this.  Classic references like Hertzberg's spectra of molecules series and Levine's book Molecular Spectroscopy also go into it in some detail. 

I think that the effects of temperature and pressure are getting confused here. 

Temperature has two effects.  First, there is a Gaussian broadening effect due to the relative motion of emmiting (or absorbing) molecule and the observer.  Secondly, the temperature determines the population of the rotational and vibrational states via the partition function.  The states are described by a vibrational quantum number n and rotational quantum number j. Except at high temperatures only n=0 and low j numbers have significant population.  The symmetry of the molecule determines which delta j's are allowed.  While the initial population distribution is significant for emission, it does not matter much for absorption (except at very high photon densities where there is a photobleaching effect) because all allowed transitions are available. 

Pressure:  All quantum states in molecules and atoms have an associated natural lifetime.  A longer lifetime gives a greater certainty in the energy of the state (Heisenberg Uncertainty Principle) and hence a narrower line width.  The effect of collisions is to perturb state and shorten the lifetime and hence increase the uncertianty in energy.  This leads to a Lorentzian broadening.  gncp was describing the effect of mass on the perturbation.