Don't think about your panels as being 4 ohm panels and 8 ohm panels, because they aren't. You have to wire them according to how they actually measure. Otherwise, you'll end up with a grouping with either a much higher than 8 ohm impedance, or an unusually low impedance group. Especially since some of your panels measure at 2.7 ohms. Plus, your crossover will have to be calculated using actual measured impedance's. Otherwise your crossover points/slopes won't be accurate.
And speaking of the crossover, any ideas? First order, 2nd order, or greater? Parallel or series? Crossover points? Active? Brian, at VMPS, used 6db first order series crossovers. His bandpass points for the panels were 280Hz, and around 7.5KHz, since the panels are fairly full-range. The panels go higher, above 15Khz, but they'll get a bit beamy if used that high. Plus, there's a slight built in response peak at 12KHz some users like to avoid by using a bandpass design, then handing off to separate tweeter(s). Though you could add a zobel circuit to eq the peak a little flatter, if desired.
Finally, besides having panels of 2 different impedance's, you have panels of 2 different styles. In your picture, the top and bottom panel, in each stack have 13 aperture openings (with curvy fluted edges) (4 total), while the other 16 have 8 aperture openings. So, I'd recommend using those 13 aperture panels together as the middle 2 of a column, since those 13 aperture panels have a much broader dispersion pattern. Then use the remaining impedance matched 8 aperture panels above and below the center pair. You could have a 3+2+3 layout, or a 4+2+4 layout, depending on how your individual panels impedance's match up. But, the center 2 panels combined impedance should be slightly lower or equal to the other panels, so they aren't too low in level, compared to the others.
Hope this makes sense!?!