While the specifics are somewhat brand-dependent, this question has essentially been answered already in one of your follow-up questions.

Start with the following assumptions:

1. There is no magic involved; everything that happens will be as simple as it possibly can be and still work;

2. The system is not and cannot be foolproof; any sufficiently advanced fool can easily defeat it (a corollary is that the system can be "gamed" to advantage by anyone who knows the system);

3. Modern TTL systems, unlike the TTL-OTF systems used in the film era, do not measure the actual exposure while it is happening; and

4. In an optical multiflash system, no information can be transferred from flash (or IR controller) to flash without something flashing. (Radio systems, like the new Canon 600EX-RT/ST-E3-RT can do some additional trickery without flashing, but there is no evidence to suggest that the method of TTL metering is different.)

All of the remote flashes are fired at a known power level (1/32 power usually, but that's brand-dependent) at the same time, before the exposure, at the command of the "master" in the system. This will probably not result in a correct exposure; the metering system will decide how much the overall flash power needs to be adjusted (and in which direction) in order to achieve a "correct" exposure. This is the "0.0", uncompensated power.

If you have made no adjustments, either overall using flash exposure compensation (or, in Nikon's case, the combination of exposure compensation and flash exposure compensation) or using the group settings, that is the power level that will be relayed to all of the flashes in the system (all the flashes "listening" on the same channel) before the main firing signal. If you have set an overall compensation, then the "master power level" (the "0.0" value arrived at above) will be adjusted appropriately before being relayed.

If you are using groups set to different power levels, or the "ratio" feature that's sometimes available (particularly in macro systems), then the "master power level", adjusted for overall compensation, will again be adjusted in accordance with the settings you've used. If you have Group B set to +1.0, then all of the flashes in Group B will be told to fire at a power level that's twice as high as the compensated "0.0" value. If Group B is set to -1.0, of if Group B is on the "1" side of a 2:1 ratio setting, the flashes in Group B will be told to fire at half the power level.

It is, of course, possible for all of this to go horribly wrong, and the chances of it going badly increase with the number of flashes you use—if you don't understand how the system works for your camera and flash system. If you do understand the system, you can do an awful lot more than the limited number of groups and settings available may suggest.

You need to know what your camera is metering. You need to understand that a bare speedlight close to the thing it is illuminating will have a much greater impact on the metering exposure than a diffused speedlight at a greater distance. You need to understand that, say, a 580EX will put out more light at a given power level than a 430EX will (the same goes for the SB910 versus the SB700 and similar pairings), and that it's the relative power levels, not the absolute output, that the system is adjusting. And that two speedlights in one softbox or umbrella will put out twice as much light as one speedlight in another at the same power level (so that you can still get a 2:1 lighting ratio even when all three flashes are in the same group). And that you can gel individual lights down within a group as well to reduce the power of individual flashes without using up an extra channel.