So basic structures remain but change, whereas transitional structures (worldviews) are replaced. E.g., postmetaphysics replaces core assumptions about metaphysics. You can see this in the other section of the article on Deleuze and DeLanda quoted above:
"It is a central concern for Deleuze...to do away with...ideas about ‘timeless forms’ or ‘essences’ that emanate from some Platonic heaven to give shape to the world of real things. Deleuze finds that such ‘essentialism’ pervades our normal perception and ways of thinking...things are thought of as belonging to categories and sub-categories which are defined in terms of invariant properties or, again, essences."
Also in that section he equated Deleuze's virtual realm with Piaget's primordial and undifferentiated continuum. Deleuzes' virtual though, given the above, is not some kind of Platonic essence, since it took can be and is affected by the actual in a reciprocal process.
The author at the end of this section tries to argue though that Deleuze does allow for increasing complexity in an enveloping sequence, that the real including both the virtual and the actual have this tendency. An earlier quote above notes that Deleuze refutes evolutionism and levels, and sees it more as strata and lines of flight. I'd suggest that this too is the difference between the postmetaphysical and metaphysical worldviews, which replaces predecessors but includes previous but transformed structural elements.
I'm also reminded of Luhmann's notion of each 'level' remains an autonomous system that structurally couples with the others. Indeed the earlier levels set up embodied parameters for later levels while themselves undergoing further developmental transformations due to their coupling with later levels. But it's a different (postmeta or popomo) orientation to how the process is structured.
But he keeps trying and makes some good points on how space is structured by bodily image schema. He doesn't use that term but by his descriptions that is to what he is referring. Deleuze didn't make reference to such schema, at least in terms of cognitive science. And that frame could indeed expand on his ideas, as he did a few posts earlier showing how the fold relates earlier image schema with the most advanced math.