As Jeff mentions, once mixed with the SBD you may find the overly side-heavy balance of the on-stage mic pair is no longer an issue. With a "two musicians trading songs with each other" type of setup I feel it's okay to have a each performer positioned to one side in the stereo panorama, yet by how much, and in what way, is of course a subjective call.
If simply mixing the two sources isn't sufficient, the easy out is simply panning both 4023 channels slightly towards center (or both SBD channels if that is in stereo) until it is sufficiently balanced. The problem is this will also narrow the room sound and limit the immersive feeling. If the SBD has stereo reverb on it, determine which of your two sources has the more spacious sounding reverberation and leave that source panned as wide as possible, panning the other source toward to center to do the balancing.
IME, this is easier to work when there are sufficient early-reflections and ambient reverberant room sound in the "unpopulated" side, such that even though the performer is heard as being located far left (or vice-versa) in the stereo panorama, there is still plenty of acoustic energy across the entire sound stage - such that in the acoustic power sense rather than the psychoacoustic balance sense of perceived performer position, the right speaker is working nearly as hard as the left.
Panning the on-source-axis microphone somewhat toward center (while simultaneously increasing the level of the other off-source-axis mic which is not-panned toward center - the one facing the non-active performer) is one way of pulling more of this from your existing recording without compromising the overall stereo ambient width as much. This is basically the same as what Jeff is suggesting, with the addition of increasing the level of ambient energy in the opposite channel, helping to retain sufficient reverberant width without making everything overly mono-y. Advanced Mid/Side re-balancing can be useful for this. Either way, you'd need to dial in the appropriate correction for each performer and switch back and forth between the two as they trade songs.
Use of omnis in place of the cardioids would sort of automatically do that for you. Better, a separate room-sound microphone pair or even a single channel of room/audience ambience (with only limited direct sound in it) in combination with your 4023 pair, is another more controllable way of achieving this.
Unfortunately I don't have editing software loaded currently, else I'd help out.