This is rather belated, but it speaks to a question I wrestled with and have (so far) not been able to get NOAA to address. But I think I can add something to the discussion.
The NetCDF files throught June 27, 2017, were Stage III files and contained no projection information. In the archive folder, where all the pre-06-28-2017 files are stored, there is a small C-code program that lets you obtain lat/lon from the [y,x] grid points. It also has a shapefile (allpoints.xxx) that contains HRAPy,HRAPx,lat,lon values. Starting on June 28,2017, the NetCDF files are Stage IV, and do contain projection strings. However, the Stage IV files use a larger grid, and no longer contain the offsets mentioned above. What I wanted to do was obtain precip data for each gridpoint, and you cannot use the archived program (directly) to convert the grid points to lat/lon with this dataset. But you CAN modify the archived code slightly to do this. The Stage IV files have values for each y and x. These are the easting/northing values for each gridpoint. They take into account the offsets that had to be applied in the Stage III version. In order to use the embedded y,x values, you need to divide them by the mesh length (4762.5 m) in order to obtain a true grid value, and pass that to the HrapToLatLong() function. FWIW, the Stage IV NetCDFs can be properly displayed by QGIS -- just select it using the option to add Mesh datasets.