I should find the link, but do a web search for the Walter Lewin physics lecture about why the sky is blue and clouds are white.
He explains and demonstrates why smoke consisting of extremely small particles scatters primarily blue light (short wavelengths) while smoke or particles that are larger scatter all wavelengths (white light).
There is a threshold of particle size that determines the type of scattering you get. Rayleigh scattering versus Mie scattering.
Clouds look white due to the large particles causing Mie scattering. The sky looks blue due to Rayleigh scattering.
The point is: When smoke looks white, you know that it consists of much larger particles / droplets, but when it appears blue, you know that the particle size is far smaller.
From what people have observed, what you want for the best flavor are only the tiny smoke particles. The larger ones consist of more bad-tasting compounds, creosote, etc.
I'm speculating here, but:
The challenge for us is to create conditions of combustion that generate only the blue smoke, and/or condense out the larger particles before it reaches the food in the smoker.
I know this is off topic, but I wonder if what we want is a situation where the smoker fuel (pellets, dust, chips, etc) burns, but not too well. Enough oxygen to let it smolder, but not so much that it can burn too hot or something like that. It might be that as we strive to keep our smoke generators from going out, we can go too far. And what we really want is to maintain a fine balance between failing to burn vesus burning too well.
And that may be why we relish the effect when a smoke generator burns completely, but burns for a long time (not too fast).
And that means that we always need to adjust and maintain conditions. And it can be different for different days, wood types, wood form (pellets, dust, chips, chunks), generators, weather conditions, smoker loading (how much, and how moist the meat is), etc.
This may be part of why a mailbox mod or the like is helpful. You can control some of the variables of combustion more easily by separating the combustion from the smoking chamber conditions. Plus, you get the chance to condense out the larger particles before they get to the smoking chamber, which makes it even more forgiving. You can lean toward more positive combustion (for reliability), but condense out some of the bad smoke constituents that result from that "too good" combustion.
Again, I'm just speculating here about some of that. But I want to set up a smoker system that lets me experiment with some of this!