I must've overlooked this thread until today. Its right up my alley. I do a lot of suuround recording and I agree with most of what has been posted previously. Following are a few thoughts and clarifications from my experience-
First bear with me through a breif philosophical underpinning, as I think a rudemenatry inderstanding at that level really helps to point the way at each practical step.
I consider live music performance and environmental ambient recording, via either 2-channel stereo or multichannel surround, as all being rooted in capturing the sound arriving directly from the sourc(es) and the indirect diffuse sound that arrives from all directions. Those are the basic things that must be in balance. With 2-channel recording we generally arrange things to achieve a good set balance of those things via the recording location, microphone pickup pattern, stereo mic'ing cofiguration, etc. With a multichannel microphone arrangement, we can set things up in ways that achieve a less compromised pickup of each of those aspects and allow a degree of freedom over how they are combined for 2-channel reproduction, or manipulated for discrete multichannel playback, after the initial recording was made.
Practical things-
I find good capture of surrounding ambience generally requires space between the microphones, and that the direct sound component from each individual source is not being picked up by all microphones (there is minimal direct sound in at least some of the other channels). That means some spacing combined with some directionality imparted by microphone directivity or baffling.
Ambisonic mics are compact and convenient and a 1st order ambisonic microphone is great at capturing direct sound from all directions and retaining directional info. But they are not great at conveying the diffuse ambience that gives one the distictive realistic feeling of "being there" immersed in the sound environment. You need some space between the microphones for that. Sure folks use them for ambient recording, but that is mostly about convenience and/or VR compatibility, rather than representing an optimal approach
The IRT cross is a great starting point for that and what I'd recommend if using 4 cardiods or supercards for the type of ambient recording you are doing. If using 4 omnis, use a similar spacing to IRT cross with baffles to produce directionality. Mounting to the surfaces of a box works well. Mounting to four sides of your torso does too. A cross-shaped intersecting Jecklin disk will work, as will two of them separated by a fore/aft baffle. They may provide sharper direct sound imaging, but a box is simpler and effective. You essentially want sufficient directionality/spacing between each adjacent pair of microphones similar to that of a 2-channel recording atrangement. That means any two opposing channels in a 4-channel ambient pickup arrangement is angled twice as wide and spaced twice as much as a typical 2-channel arrangement. With that in mind, using 4 omnis, I'd use a box before choosing a sufficiently large beachball sized spherical or vertical round column baffle, but use what you have available.
For ambient recordings an L,R,Ls,Rs (quad format) recording and reproduction works well. For things like music where there is a definite forward direction I prefer to turn that 45 degrees (L,C,R,S), which achieves better front/back differentation, imaging across the front, and keeps the loud front arriving sound in the front. For LRCS playback it helps to distribute the S channel to as many surround speakers as possible.