Additional thought:
If we just think one step further, an implementation into voronoi would make sense, too,
for another reason regarding textured meshes:
The movement of border edges will always require to reproject the texture, because it’s deformed (= texture distorted) there.
If the operation is happening in one operator, then the voronoi cutting can happen only at the end, after all computation of edges is done within that operator.
And doing only cutting means no change for texture appearance.
So the voronoi would not cut and then displace thereafter, it would at first compute the final position of the already displaced „noisy subdivided“ final cutting edges, and then only once cut there.
The displacement computation for inner faces of the voronoi seems to me even easier to define:
It’s driven by the same global noise attribute, only constraining the „movement“ (or edge placement then) to the original surface.
If we just think one step further, an implementation into voronoi would make sense, too,
for another reason regarding textured meshes:
The movement of border edges will always require to reproject the texture, because it’s deformed (= texture distorted) there.
If the operation is happening in one operator, then the voronoi cutting can happen only at the end, after all computation of edges is done within that operator.
And doing only cutting means no change for texture appearance.
So the voronoi would not cut and then displace thereafter, it would at first compute the final position of the already displaced „noisy subdivided“ final cutting edges, and then only once cut there.
The displacement computation for inner faces of the voronoi seems to me even easier to define:
It’s driven by the same global noise attribute, only constraining the „movement“ (or edge placement then) to the original surface.