It's interesting to read about your various compression habits, I share some of the same experiences. I expected that I would slowly turn less leaned towards compression as a way of improving the final product, but over the years in my case it has rather been the opposite, right now I find compression to be really essential, I pretty much compress everything. Multiband compression is something I'm currently looking into a lot. The reason is that I like a lot of compression, but not when it damages the stereo image. I'm sure you know that when you place a limiter on the mix bus and just limit everything a lot, there is a lot of frequency fighting on each speaker and hence the stereo image to some degree collapses. How I counter that is instead of limiting on the mix bus scope I apply multiband compression on non-dominant frequencies on each sound source - in parallel. This essentially does the exact opposite to heavy mix bus limiting - since it reduces the RMS on each speaker it also enables more unique peaks from each sound source on each speaker when you compensate for the RMS loss - the sounds on each speaker fight less and therefore the stereo image is much improved. This is one of the pleasant characters I focus on creating when I mix. By doing this especially on the most prominent sound sources in the mix - the ones that normally consume the most mix signal - you can free up a lot of mix signal too and create air, so this kind of technique can be used both for creating the air characteristics (by not compensating so much for the RMS loss) and also for creating more rich stereo characteristics (by compensating for the RMS loss). This is an example of how you can work with compression in mixing and get multiple different characteristics out of it, depending on what you do with it. Another example would be to create rhythm characteristics from it, by setting a long release time. In this case, to add rhythm you do not increase the volume on this bus, bur rather the opposite. So what you can do here and I do this a lot, is to set the release times rather long going into the first verse and then gradually reducing it on a cumulative curvature to make the song groove. In other words, it is really important to think of a volume fader of an effect as causing a desired effect not only by increasing volume/wetness, but also when decreasing volume/wetness. Essentially every side-effect/bi-product that you notice provides a certain negative characteristiics, can be used to create the opposite characteristics. This you might find inefficient, but in the context of the mix it can help create interesting characteristics. In mixing, by removing you sometimes add and by adding you sometimes remove. Almost everything you do in mixing has multiple impacts, it adds something and removes something else.