Captain Whiskey has expressed a desire to actually understand this issue. It's apparently not well understood by most, because it's a fairly complex phenomenon. Frankly, it pretty quickly becomes so complex that it's not worth trying to think through. Your ears can "do the calculations" a lot faster than your brain can. That's part of why I said to just listen. Another part is that it really doesn't matter what it looks like, or what our cocktail napkin figuring tells us. All that matters is how it sounds.
It is completely meaningless to talk about the kind of thing in the first post in terms of phase angle or "degrees". It's just a futile effort and will just about always start exactly this kind of confusing argument of phase vs polarity. If we need to put numbers to it, then we can talk about delay time in terms of seconds or milliseconds or samples. We can talk about distance in feet or inches or centimeters or whatever you prefer. But the phase angle is dependent both on that delay and frequency, on both distance and wavelength.
If we take a slightly out-of-tune A at 441 Hz, it takes 100 samples (at 44.1K) to go through one full cycle. A delay of 50 samples will put that note out of phase by 180 degrees, and if mixed back with the original, it will cancel almost completely. But if you've played that A on a guitar, you're not just hearing that one frequency. You'll also have some amount of its octave - 882Hz, which has a period of 50 samples. When you slide that track, both the 441 and the 882 go the same 50 samples right? So the 882Hz has been shifted a full 360 degrees, and will add between the two tracks. Every other frequency in the spectrum will have some other phase angle, and reinforce or cancel more or less. You end up with peaks and dips in the frequency spectrum that look like a comb and really have a pretty distinctive sound to them. It's very much like the sound of a guitarist's phaser pedal if it stopped sweeping.
Now, the depth of those peaks and dips depends on the "coherence" between the two tracks. It will be "worst" when you have exactly the same thing on both tracks (say, a digital copy, or a parallel send path with uncompensated latency) and mix them at exactly the same level. Without any delay, if you put two copies of a track side by side, mix them to exactly the same level, and then flip polarity of one, you get silence. Do it now. You have all the tools right there in front of you. Try it! If you play around a bit, you will find that as you change the volume of either of the two tracks - whether up or down - the original sound starts to come back up in volume. It's really as easy as this: |1 + (-1)| = 0, |1 + (-1/2)| = 1/2 = |1/2 + (-1)|
Now if you delay that duplicate track by a bit, and mix them up to exactly the same level, you'll probably find that most of the signal goes away, and you end up with some version of boxy/washy/phasey treble content. If you mess with the faders you'll find that it pretty quickly returns to sounding "normal" as one signal gets louder than the other. Try it with a white noise track and look at the mix output on a spectrum analyzer. You'll see the comb when the levels are the same, and you'll watch it level out as you change the relative levels. You can also watch the peaks and dips move as you slide the one track around in time. You might also notice that peaks turn to dips when you flip the polarity switch on one of the tracks. It's great fun! Play with it, you can't possibly hurt anything as long as you keep your listening levels reasonable.
Now, so far we've kept the two tracks exactly the same except for volume, polarity, and delay. Most of the time, though, the whole point is that the two tracks are noticeably different. One mic is further away, capturing a slightly different frequency spectrum, maybe a bit more room ambience. Or you compress the fuck out of one of the mics, or send it through a reverb, or... You'll only really hear the comb filtering in the parts of the signal which are still the same between two tracks. Go back to your two duplicate tracks, all nicely time aligned, mixed to exactly the same level with one polarity flipped. Now drop any plugin you like on one of the tracks. Pray that your plugin latency compensation actually works and try to keep approximately unity gain for the bulk of the signal. Now mess around with the controls on your plugin and watch and listen as the differences created by the plug pop out! This part of the signal will not be affected by any comb-filtering caused by delay or distance. Slide one of the tracks around and see what happens.
So, I just typed a bunch, but don't know that I've given a lot of new information in the actual text. If you played along at home, then I think you will have learned quite a bit.
But you asked for dos and donts.
Do - whatever it takes to get the sound that you're looking for, no matter what anybody says is "right".
Don't - worry about "problems" which don't impact the outcome.
Do - experiment and explore any topic which interests you until you are satisfied with your understanding.
Don't - get so caught up in theory and knowledge that it boxes you in or limits your options.
Do - When micing one single source with more than one mic, try to decide on what specific role you want each mic to play, and try to keep them from overlapping too much. One is for the direct sound, and the other for the room. One is for the low end and the other for the treble. Etc.
