A couple important notes-
The comb-filtering seen above is what happens when the two signals are combined electrically. In other words, mixed together to form a single monophonic signal. This is the basis upon which 2-channel stereo recorded using spaced omnis is considered to be less than optimally mono-compatible. Of course, in reality a recording made using spaced omnis is perfectly mono compatible by simply eliminating one channel and only using only the other- "wa-la" we then have a single monophonic omnidirectional channel with zero comb-filtering, from a single microphone placed just slightly off the original center-line.
That clear and distinct comb filtering does not manifest in the same way when the recording is played back normally in 2-channel stereo. It will not manifest at all with headphone listening since the signals don't mix at all until they are in the listener's brain. With loudspeaker playback, the two signals mix only upon reaching the listeners ear(s), the sharp nulls go away, and the comb-interaction will be different at each place in the 3-dimensional space of the listening room, as well as being different for each frequency in question. Regardless of the specific location of the listener's ears in the room relative to the speakers, there will be no sharply defined comb cancellation notches as seen in the graphs above, that only happens when the two signals are mixed electrically. At anything above the first notch (which shifts around by both frequency and position as noted) the combing interactions become so complex they essentially become random.
There are going to be level differences in most widely spaced omni setups. Any source which is close enough to the mics and not on the centerline is going to produce level differences as well as arrival time differences in the two channels.
For spaced omnis out in the audience, the level differences will be mostly nearby audience sounds, and one reason I like wide-spaced omnis of 5' or 6' is that the sounds from nearby audience members will not be recorded with the same level in each channel unless they are directly in front or in back of the mics. Because of that, those "unfortunately too close and unwanted" sounds image off to one side or the other, "getting out of the way" of the music imaging across the center between the two speakers. It's still there, but becomes less distracting and offensive, easier to mentally separate from the music and ignore.
For wide-spaced omnis placed at the stage-lip or on-stage, the increased proximity of the various sound sources on stage to one or the other of the two microphones will produce level differences in addition to time of arrival differences.
Even quite wide-space omnis isn't going to decorrelate the lowest frequency information. The spacing may be wide enough at those frequencies to produce some stereo difference phase information, but to be decorrelated, the signals need to have a difference of more than one full wavelength. The first cancellation notch in the graphs appears at a frequency where the phase difference between the two signals is exactly half a wavelength apart (the phase difference = 180 degrees).