As long as your DAW/plugin correctly apply latency compensation you should not have any phase issues.

This might sound like I'm being an asshole so let me just say I don't intend it as such at all:

Use your ears not your eyes.

Sharp spikes at a particular frequency is the nature of resonance, so of course it will look that way on a visual frequency analyzer. But ignore the analyzer and listen, does it sound right? Does it feel like it's fitting in the mix? If yes, then it's fine. If no, then there is work to be done but base it on what you're hearing not what you're seeing.

IMO analyzers are great for a quick reference point for very broad strokes, but are next to useless when it comes to narrow bands within a particular mix.

All that said, maybe post a screenshot to show just how much of a spike we're talking about? Include an audio clip if we want to get really serious about it. Discussing audio is all fine and dandy but it's pretty futile without all hearing the same thing.

Bear in mind that's only summing, the differences become more noticeable with additional processing, but you're absolutely right: as a hobbyist it's not worth the money to invest in that type of gear yourself. As a professional studio charging for services then it is, and you as the hobbyist can always pay a small amount to that professional studio for mixing or mastering and get the benefits without having to buy the gear yourself.

Here's a blind test for you, this is PRE mastering, no use of analogue in the mix except for summing. One has been summed digitally, one has had the main subgroups summed in analogue (using an SPL MixDream XP).

A: https://drive.google.com/file/d/0B_ptf08bafFwTjhaNGF6bU1yUUk/view?usp=sharing
B: https://drive.google.com/file/d/0B_ptf08bafFwNFpsZUtWdUpib2M/view?usp=sharing

I know I hear the difference, and the two definitely do not phase cancel, not by a long shot. Which is which?

How loud are we talking about exactly? I try to do most of the critical work around 85dB SPL which is the level we hear relatively flat at, but even that is pretty 'loud' to me for long periods so I turn it down in between.

Taking frequent breaks helps, get up and do a lap around the house in silence. Continually drink water or tea so that you are forced to get up and go to the bathroom frequently (plus it's good for you!).

A lot of things can play into ear fatigue though, not just the sheer volume. Your monitors will play a huge part, if they tend to be fairly mid-forward and present that is going to be more fatiguing to a lot of people. And similarly, how well you're actually mixing will have an impact too. If your mix is full of overlapping elements fighting for space in the midrange and being squashed by compressors and limiters, that right there is going to get very fatiguing very quickly, and not just to you but also to the listener at the end of it... another reason to not participate to heavily in the loudness wars.

The detection circuit for 'glue' I just mean that it's likely designed to react to the averaged signal instead of the peak signal. So things like kicks and snares will pass through relatively untouched and instead the compressor will react to the slower-moving changes in the track's composition.

This is why it can be usefuly to have more than one type of compressor. Something that works more on an RMS level and another that works more on a peak level, each in smaller amounts than you might do if you were only using one. They're doing two different things and can in turn get a more transparent or musical result, but it all boils down to taste really. Experiment and find out what works for each particular track, now that you've got a better understanding of how they work it should be easier to apply them in different ways to get a particular result.

Yep you got it.

I oversimplified it slightly, but that's the gist of it. With the longer release times it the gain reduction will still be varying, but it won't release as much, or release back to 0 as often. So it wouldn't just be pinned at -3 GR constantly, but it might fluctuate between -2 and -3 instead of between 0 and -3. You'll get some level increase, but beyond a certain point it will stop getting louder (until you hit a quiet section and it releases and sounds really weird).

And yes for that extreme volume gain the tradeoff is potential pumping, you want to try to time the release to the music so that it's at least pumping in a musical fashion. Or you can go with an near-instantaneous release time (i.e 1ms) which won't necessarily pump in the same fashion but at that point it becomes more like a clipper and begins to distort.

Answer 1: The faster the release time the quicker the full amount of make-up gain will be applied. Sort of. A slow release time will 'smooth' as you say, keeping a somewhat constant gain reduction in effect. A faster release time will continually being resetting back to 0 GR so all the bits in between loud transients are getting boosted, and overall it appears louder. With that said, the trade-off for quicker release times is that it becomes more prone to pumping, smearing, and distortion.

Answer 2: A regular stereo compressor doesn't care what frequencies it's working on, it's purely a dynamic processor so it will affect all frequencies all at once. In the same fashion as the limiter (a limiter is just a high-ratio compressor, in it's simplest form) it pushes peaks down, creating headroom so that the overall volume can be increased. It makes things sound 'loud' by reducing the gap between high peaks and low peaks. A multiband compressor does the same thing but it first splits the signal into multiple bands, and applies a compressor to each of those split bands, so each range can have it's own compression settings, unaffected by the other bands.

Answer 3: Partially answered above. As for is it 'better'... yes and no. Usually no, but that's my opinion coming from the mastering perspective, and many will argue the opposite. To me multiband is a surgical tool that should be used when everything else has failed, generally if I'm having to put a multiband comp on a master it means that something has gone horribly wrong in the mix. If you have access to each instrument, fix the problem there. Generally compression is something done to tie things together, so when you get into splitting them apart even further with a multiband things can get really messy.

Answer 3.5 (you have two question 3s!): While I said I don't like multibands on masters, they do occasionally have execellent uses in the mix. My main example would be on a distorted bassline (think neurofunk-style drum & bass) that's a bit all over the place. A multiband comp would let you tame the sub range to keep it at a constant level while still allowing the upper harmonics to have dynamics and movement.

Answer 4: "Glue" is just the term used when you're using compression to get that feel of cohesion between all of the elements, making the track feel like one thing instead of obvious layers that don't quite fit together. Often what it means at a technical level is slower RMS detection circuits instead of peak.

Answer 5: You're forgetting the element of TIME! When the kick hits, the compressor will act on everything (including the pad) pushing it all down. The kick will be finished quickly, but the pad is long so you'll hear it gradually rise back to volume as the compressor releases. So they'll never be at the same volume at the same time... they interact. But also think of the nature of a kick vs a pad, a loud transient vs a steady tone... you wouldn't WANT those to be at the same level, or the pad would sound insanely loud and the kick would be tiny. Balance does not mean everything is the same, balance means everything is comfortable relative to each other.

Answer 6: Simplest version - when you want to change the dynamic range, not the volume. Reach for the fader first, if you're finding that an instrument is standing out too much in some spots and then getting buried in others, that's when it's time to reach for the compressor.