Normally this is used when starting a board, by selecting all subcircuits and then dispersing them. This scatters the subcircuits around the board so that you can pick individual ones, rather than have all the subcircuits at the same 0,0 coordinate and thus impossible to choose from.
When the first argument is "buffer", rearrange subcircuits within the buffer so they form a tightly packed rectange centered at the buffer origin. This is useful after the initial import schematics: from sch-rnd cross-select a bunch of subcircuits making up a functional block of the design (e.g. a PSU section), cut to buffer, DisperseElements(buffer) then paste at a spot of the empty board.