Think of this simply as an advanced, single-point coincident stereo mic with additional adjustability. You tune the mic pattern and angles in post to optimize the situation as best as you can given your mic placement, but you can't move the mic location and there is only so much directionality available with any first order mic pattern to begin with (omni thru cardioid thru fi-8). One of the intersting things for me in playing with decoding Tetramic recordings has been getting a better experiential, 'seat-of-the-pants' feel for that range 1st order directionality, given the coincident restriction.
Think about it- the 1st order pattern with maximum rejection to its rear is a cardioid, and to get stereo with a coincident array you need to angle the two patterns apart. But the more you angle them, the more you compromise the rearward rejection. So like most things, dialing in the best decode often entails a number of compromises. I quickly started thinking about how nice it would be to actually have something like 3rd or 5th order capability.
Forget about the capsules that point to the back. Or rather, think of all four capsules as either adding or subtracting sound as determined by the complex matrixing that the decoding software does to synthesize the virtual microphone patterns and angles. In that light, the rear capsules are doing their job to reduce sound to the rear of the mic, if that's how you choose to decode it, given the limitations of 1st order, coincident microphone patterns.
IMO, The best solution for the specific scenario you describe above might be seperate cardioids facing more or less straight ahead to achieve maximum rejection of the crowd, but spaced apart like an A-B spaced omni setup, which would move each mic slightly away from the center of the drum kit and closer to the guitar and bass. I often do that with 3 or 4 omnis across the stage and the proximity of each mic to the instument sources is enough to make the level of crowd noise acceptable as 'live ambience'. However, that's a configuration that is unavailable with a single coincident mic, ambisonic or otherwise, and pretty much the same thing applies to any near-coincident setup as well.
I have a few Tetramic recordings of Bobby Lee Rodgers in a jazz trio setting that were made in exactly the layout you describe, same proportions and angles just minus the rowdy crowd members- guitar amp(s) to the left, drums center (unamplified), acoustic bass right (bass amped through two small monitors arranged farther back on either side of the drum kit). The Tetramic was placed just-under/in-front-of our front-center high-top table about 4'-5' from the drums and about 18" to 30" above the floor depending on the recording date. Once I get the recordings together and converted from A to B format, I'll work on a way to share both the stereo mix-down I arrived at as well as the B-format files with anyone here who is interested, so that you can play around with the decoding yourselves using any of the the available B-format plugins.
I've been intending to make something like this available, but it will take some time to get together.