Pair that to a x,y stepper and some Python OpenCV script and you have a ready to go "cell counter", which this paper espouses as one of its uses. Not only that but they only get 20nm with shadow imaging techniques which (I'd imagine[2])are inherently awful for actual examination of features.
Here's[3] a paper that uses wet-methods (i.e. good for biology, where you can add florescence/dyes/stains to your culture; not nearly good enough for modern chip fabrication) that gets you down to the 20nm range. And at that point your lab already has a microscope to hit those lengths.
Basically this lowers the price of this use-case: "oh at t(0) we have 30 blurry units we can sorta-kinda safely assume as eukaryotic; t(1hr) = 600 blurry features". I can see this being a market opportunity for having in-house cancer biopsy diagnostics.
From a research point of view, you're doing at the lower level (protein folding, crystallography, whatever), you're already equipped with the proper mass-specs/NMR/SEMs/TEMs you picked up from Waters or Agilent at half a million. I mean that's got the potential to save a lot of lives, which is awesome, but I probably sound disappointed because I was really hoping for an actual advancement in lens-less microscopy.
--
[1] en.wikipedia.org/wiki/Diffraction-limited_system ~200 nm under ideal conditions, no pedantry please on apertures or practical things like chromatic/spherical abberation issues, etc. I realize these limits exist, we're just talking orders of rough orders of magnitude. [2] Not in the field, just remarking based on conversations with people who have had to setup labs at universities with limited NIH/NSF budgets. [3] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2645564/ There are a boatload of other papers out there, but this is the one I had off hand.