I have come up with a few heuristics gleaned from the literature to help choose which cells from which libraries are the most likely to yield reliable lower VDD circuits. For really good performance on 0.2 - 0.6v range it’s necessary to have a new library with a change in transistor lengths and a look at different size ratios. It looks like we can get below the 1.2v suggested for sky130 with this focus: lower track size library (less capacitance) + x2 or x4 cells as the baseline (more fingers) + hs library variant (lower Vt). Here is a handy recent source on low voltage cell libraries: http://theses.ncl.ac.uk/jspui/bitstream/10443/4450/1/Morris%20J%202018.pdf
I will wait until the rest of the OSU cells come out to check these heuristics out but I suspect the OSU cells will work better than the factory cells at low voltages because of the choice to use metal1 instead of LI.