I have recreated my comments below. If anyone thinks they are in some way inappropriate or wants to comment why they were removed I'd be interested to hear.
I explored this question with the free magnetic F.E. software "FEMM".
Attached slide show has the results. Summary of slides is as follows:
SLIDE 1: Analytical calculation of single loop in air. Loop radius = 0.03m, Wire radius = 0.005m. L=7.05E-8 H
SLIDE 2: Same Geometry, shown in "axisymmetric" view. i.e. (r,z) plane where radius measured horizontally and height vertically. The center axis of the loop is the left side of this figure
SLIDE 3: Same Geometry (simple loop) FE solution. L = 7.86E-8 H.
SLIDE 4: Add Iron Slug. L = 4.47E-7 H
SLIDE 5: Same-polarity loop pair (no iron). L=1.03E-7H for our loop
SLIDE 6: Add iron Slug to same-polarity loop pair. L=8.2E-7 H
SLIDE 7: Opposite-polarity loop pair (no iron). L=5.4E-8 H
SLIDE 8: Add iron Slug to opposite polarity loop pair. L = 6.8E-8
SLIDE 9: 6 opposite-polarity loop pairs arranged symmetrically, no iron. L = 5.5E-8 H
SLIDE 10: Add iron Slug to 6 opposite-polarity loop pairs, L= 6.6E-8 H
SLIDE 11: 6 opposite-polarity loop pairs arranged irregularly (no iron). L=6.75E-8 H
SLIDE 12: Add iron Slug to 6 opposite-polarity loop pairs. L=1.36E-7H Significant increase.
SLIDE 13: Change iron slug from slide 12 to iron enclosure (enclosing the current.) L = 1.32E-7 H
SLIDE 14: Change iron slug from slide 4 to iron enclosure (enclosing the current.) L=4.4E-6 H
All of these are actually dc simulations. Neglecting eddy effects, they are also representative of single-phase 60hz circuits. Certain geometries are claimed to act similarly to 3-phase circuits as will be elaborated in the discussion below (under slide 7)
Slide 1 is simple analytical solution of a loop as a test case.
Slide 2 shows an overview of the axisymmetric geometry used for the FE solution. The geometry is very similar to what is used in this link, including boundary conditions described in section 2.2:
Slide 3 shows FE of the same geometry as slide 1 and the inductance is calculated... results agree reasonably well.
In slide 4 we add iron slug which causes a dramatic increase in inductance (factor of 6 increase) for this single loop, as expected.
In slide 5 we add another identical loop with current flowing in the same direction. The inductance is reported as 1.03E-7H (compared to 7.8E-8H for single loop inof slide 3). This does NOT represent the inductance of the series combination of the 2 loops (that would be at least twice slide 3). What L represents in this slide and all remaining slides is the total flux linking one specific loop (the one with the red arrow) divided by the current in that loop. So we can think of it as inductance per loop or inductance per length when we compare. The effective inductance of our loop goes up by roughly 20% from slide 3 to slide 5 when we added another loop of same polarity nearby.
In slide 6 we add an iron core to the geometry of slide 5 which results in dramatic increase in inductance (factor of 8) as expected.
In slide 7 we examine an opposite-polarity loop pair in air. * The single-phase opposite-polarity loop pair is interesting to us because the current and flux from the pair sums to zero, so we can think that it acts somewhat similar to the sum of 3-conductors in a 3-phase system (except the opposite-polarity single phase loop pair is much easier to set up, which is why I used it). The inductance (per loop) is now 5.4E-8. It is around 20% lower than our single loop of slide 3. That is similar to a set of 3-phase conductors... their effective inductance decreases when we put them close to each other.
In slide 8 we add an iron core. The increase in flux due to adding core (from slide 8 to 9) is very modest at approx 20%. This is much much lower than all previous configuraitons where we saw 600%-800% increase from adding iron. When we have our conductors arranged in groups that sum to zero (like our opposite-polarity pair or a set of 3-phase), the iron tends to have much much less effect.
Slides 9 and 10 resemble slides 8 and 9 except we have used 6 symmetrically-arranged opposite-polarity loop pairs (resembling 6 symmetrically arranged 3-conductor cables). There is very little change in inductance from adding additional pairs if the symmetry is maintained.
Slides 11 and 12 resemble slides 9 and 10 except the pairs are not symmetric, they are placed in a haphazard arrangement. In absence of iron (slide 9), there is not much change (about 20% increase from slide 9 to 11). With iron in the picture, the asymmetry is more important (about 200% increase from slide 10 to 12).
Slides 13 and 14 resembles slides 12 and 8 respectively, except that the iron slug (going through center axis of the conductor loops) is changed to an iron enclosure which encircles the conductor loops. The intent was to explore whether enclosing had a different effect than the iron slug For slide 14 (single loop/opposite direction), there is a huge difference.. the inductance is around 4.4E-6, which is more than 6 times as high at the same geometry with slug (slide 8). For slide 13 (6 opposite-polarity loop pairs irregularly arranged), the inductance is 1.32E-7 which is roughly the same as the same geometry with slug (slide 12). We conclude that closing of the iron loop makes a huge difference when the enclosed electrical circuit has non-zero net current, but not much difference when the electrical circuit has zero net current (like a 3-phase circuit).
Interesting to note that iron alone (slide 10) and assymetry alone (slide 11) did not have much effect, but with both present there was a more dramatic increase as in slide 12(sort of a synergistic effect). If we consider slide 7 as represntative of straight run of cable (including 2 opposite single phases or 3 3-phase wires), then the "worst case" that we created by looping around iron and adding assymetry (slide 12) is slightly more than twice as much. For other patterns of assymetry it certainly may be more than that factor, but my gut feel is as others said that it is not a tremendous problem unless you were very marginal to begin with.
You were right to ask about asymmetry though. That turns out to be an important factor.
Other miscellaneous notes:
All conductors are the same size and the target coil is always the same radius = same distance from the axis on the left... if they appear different it is just because I was sloppy in zooming to different levels for the screen shot.
=====================================
(2B)+(2B)' ?
