anyone seen this with openram before?

I believe the problem is that the abstract view of the SRAM (likewise for all macros) is used when routing in Openlane, and the abstract view is made by dropping large obstruction layers over the whole cell to keep it from getting routed over. These obstruction layers are kept simple---they're just big rectangles over the whole cell with cut-outs for pins, and so they often extend beyond the extents of the actual data. This means that once you replace the abstract view with the actual GDS data, you'll often find that there's nothing there (as appears to be the case in the screenshot you posted). But the underlying issue, as far as I can tell, is that the router does not honor the "wide spacing" rule for metal, or at least does not consider large areas of obstruction to be applicable to the "wide spacing" rule, as it ought to.