Right - yeah; I don't know, I had the same thought when I read the commentary (I read the paper first). Nothing in the papers suggests the ability to monitor formation/loss in realtime I don't think? My guess is this is an interpretation of the data from the fact that you see lower levels in G1, higher levels in G1/S and lower levels in early S - i.e. they must be 'transient' because the levels go up and down again.

Seeing this happen in real cells in realtime would be – I would have thought – technically almost impossible. We're at the cusp of viewing the formation/loss of clusters of RNA POL or mediator clusters with the most advanced super-res (see Ibrahim Cissé's work) but these are comparatively massive protein clusters, so the idea of being able to view DNA structural transitions at [effectively] single-molecule resolution where that transition involves a few nucleotides in a non-perturbative way seems like a reach.

Seems like the obvious next step is to break 'em with synonymous mutations and ask if there's any detectable phenotype.