From my perceptual point of view..
In spacing the omni mic pair in my multichannel setups - where the omnis can be spaced relatively widely without incurring a "hole in the middle problem" because there are additional directional microphone channels placed in the centered between the omnis - I find I most prefer the omnis spaced around 2 meters apart, and that's what I usually do.. because that's about the maximum by setup can support from a single stand. More spacing can work nicely but doesn't seem to be necessary and is more difficult to arrange because it requires additional stands or clamp positions along a horizontal railing to support them.
Looking at the coherency curve for a pair of omnis spaced 2m apart in Image Assistant, the approximate center point of the curve lies around 40Hz. Ok, that makes sense to me perceptually. With a 2m spacing no shuffling is needed to achieve a nice very low DFC (diffuse field correlation) all the way down to a sufficiently low frequency.
Now, the spacing I generally recommend in my posts here at TS for a pair of omnis used in typical taper situations is about 1m. The qualities I listen for from a spaced omni pair suffer rapidly as the spacing is made narrower than that, at least without the application of shuffling. More up to 2m is better but more difficult to setup for most tapers. So 1m ends up being the practical recommendation.
Looking at the coherency curve for a pair of omnis spaced 1m apart in Image Assistant, the approximate center point of the curve lies around 80Hz. Which also makes sense to me perceptually. Low enough to to achieve a sufficiently low DFC even though in perceptual terms it could be improved on with a bit more spacing.
However, it seems to me that most tapers have a hard time achieving a 1m spacing unless using miniature omnis. Most are using narrower spacingings, so let's assume half that. A 50cm omni spacing will place the center of the curve at about 160Hz, so most of the low bass information will have relatively high DFC.
The tentative conclusion is that shuffling may perceptually improve any recording made with an omni spacing of around 1m or smaller, and the corner frequency of the shuffling filter can be set lower the wider the spacing. The corner frequency may be determined by using the formula above or might just be set it by ear.
Here are screen shots of the coherency curves for 2m, 1m and 50cm:
