Catching up on this thread, I came across this-
De-Bleed, nope.
Point something like a wide cardioid straight down at the talkers around your stand. Use it as the bleed source to subtract from your main mics......
You can't go very far with it, and it's a slow offline process, but it can make a difference. Test for sound quality on a very small section before committing to any larger selection.
I expect for this to work well one will want to mic it so as to isolate the undesirable bleed (the nearby talkers) as much as possible in that channel while excluding the desirable sound as much as possible. With that in mind, rather than a subcardioid I'd try something like a supercardioid pointing 45 degrees down and to the rear, such that the least sensitive sector of its pickup-pattern faces the stage and PA.
One might imagine a vertically oriented fig-8 would be best to isolate local chatter beneath the mics while being least-sensitive to the desired sound arriving along the horizontal plane, and perhaps it will be for this specific application. However, I've found the null plane of a fig-8 is generally too narrow to effectively isolate pickup of distant sources that have significant level in comparison to a cardioid or supercard facing away from the source. The fig-8 null is deeper, especially in comparison to the rear-lobe of a super pointed directly away, but the more important attribute is how much reduction in sensitivity is achieved across a relatively wide angle. It's the reduction in
average sensitivity over a region that is more important in real world taper situations.
It is for that reason I use rear-facing supercardioids rather than cardioids (or sideways oriented 8's) for the ambient channels in my OMT rig. Even though it might seem pointing a cardioid null at the PA should maximully minimize the pickup of it, I've found that pointing the rear lobe of a supercardioid at the PA reduces pickup even more, due to the supercardioid's reduced average sensitivity across the entire rear-hemisphere in comparison to a cardioid.
Apologies for going OT on this, but it represents a very interesting, potentially powerful technique with applications beyond reduction of nearby talkers. One of the driving forces behind the design my OMT rigs is arranging things so as to limit direct-sound bleed into the ambient channels and ambient/reverberant bleed into the direct-sound channels, while working with the relatively limited pattern directivity of first-order microphones and the need for sufficient correlation between adjacent imaging microphone pairs. This technique could potentially be applied to actively manage this differentiation in useful ways, without needing additional microphone channels other than those already in the array.
Conceptually, it is akin to the DSP processing applied to the rear-facing supercardiod integrated into the Schoeps Super CMIT shotgun which further increases forward directivity over that of a typical interference tube shotgun design. I'm talking about the potential of extending this from the domain of a single microphone to the entire multichannel array.