The CMB is far less uniform than it should be. From an observational POV, we can see light coming to us from all directions radially towards us. We treat the s

The CMB is far less uniform than it should be. From an observational POV, we can see light coming to us from all directions radially towards us. We treat the speed as an essentially constant value, even though we know this is false in reality. It is a useful approximation that we use to simplify our models and calculations. The calculated speed of light through a medium is related to both the permittivity and the permeability of the medium through which the light is passing. When you get to the situation where any kind of changes occur in these values over the path you then see other thing happening related to the frequencies involved. These things make for all sorts of changes in the path and the speed over which light travels. An analogy might help here (might). Take a particle of any kind and have it move along a path with potential assistance in its travel in one direction and then return that particle along the same path in the opposite direction. Ask the question, what is the one way speed of that particle if all you have is the total time for two path traversal and you are assuming that the particle is the one that traverses the path at a constant speed. What can you tell me about its instantaneous velocity over the path? You do not know what assistance, if any, that particle has received. We make certain assumptions for which we cannot determine if they are true or not. To simplify our models, we take the view that our assumptions are NOT unreasonable, but we cannot test them. This is a normal state of affairs with relation to much of our scientific investigation. When we do find additional information that indicates that one or more of these assumption is potentially or actually false, we make modifications to our models and theories or we replace them entirely. We make much ado about GR and curved space-time and that "gravity" bends light. There is an assumption here that medium changes and hence permittivity/permeability changes are not applicable or consequential in the observed path changes. The thing here is that we cannot measure those permittivity and permeability changes and assume that the light is traveling is traveling through a perfect vacuum, which we also know is false. But it "simplifies" the model and calculations. We no longer use a geocentric model of the universe (this is effectively using Fourier Series) and we have moved over to a Heliocentric model because it is "simpler" to use. Yet, the heliocentric model is also wrong because it cannot take onto account the gravitational effects of all orbiting bodies. This gives rise to using perturbation theory. This makes the modelling much more complex and yet it too is not at all complete for it has not taken into account the motion of the sun and the orbiting bodies on the path through the galaxy. Again this gives rise to further complexity that must be taken into account. For simplistic models, we use the heliocentric model and for short term situations, we just ignore perturbation effects of these other things. We do no less in analog circuit analysis or digital circuit analysis, orbital satellite mapping, weather forecasting, river flows, and the list goes on and on. We use simpler models because they are "good enough" for what we are trying to do. But they will all fail once you get outside of the simplifying assumptions that underlay them. I return you to the aphorism popularised by George Box "All models are wrong, some are useful."