I just tried two cloned pour regions - ie same size, simply overlaid on the same layer, differing NET names.
Then I add a slice zig-zag keepout - when defining this initially, you could use a single thin line, to get the geometry right, then once that is defined & stable, add the parallel-line keepout.
Then Pour, and voila a split plane. Very quick.
To make this simpler, expanding to allowing a zero width trace could have uses ?
Such a trace would spawn pour clearance, but not plot, and not conduct.
This uses the feature of KiCad where it removes non connected copper, so both areas flood but neither actually doubles.
Functionally, works a treat, but this level of cleverness confuses DRC, which it seems does not check actual created copper, but gets confused by the overlapping outlines. (even tho no actual copper shorts result)
Maybe that can be fixed, so it checks actual created copper ?
Edit: This would need user care, in present form, as any user-oversight in missing a node, would join to the ‘wrong’ plane.
DRC that was actual copper based, would catch this case.
Edit2: I also experimented with very thin lines. That sort of works, but selection gets tricky, and Gerber still plots even at .0001 mm. At very low widths, the line vanishes in Default and OpenGL,shows in Cairo.
There could still be a use for 0 width trace, or a keepout polyline that has virtual width, to simplify & avoid parallel line editing.