I've attached the buffer shapefile as well as a clip of our results (both the time the tool worked and the time it did not). I am not able to include the raster data as that would give protected information away for the participants in our study - the raster data is just an NDVI map of a county in Wisconsin.
The area calculated for each buffer should be ~ 198000. Some are expected to be a little less than that due to the fact that we masked all water out of the NDVI layer. The time that it worked this statement held true. However, when we tried to run it again some of the calculated areas were as low as 15300 due to the amount of overlap among some of our buffers.
I used the exact same buffer shapefile both times and the exact same landsat data to create the NDVI layer. The only thing that I can foresee as being different is the workspace and I didn't change any geoprocessing environment variables. Let me know if there's any other information I could give that might be useful!