I've discovered a weird phenomenon using zonal statistics with a colleague at another office. We are attempting to use the Zonal Statistics as Table tool to sum the value of raster cells within polygon zones. We are using two zone polygon files: one with about 12 "discrete" zone polygons, and the other with a single "broad" zone polygon.
The odd result is this: While we get the same results running the raster and discrete zones polygon file, we get different results when running the raster and broad zone polygon file - by an order of magnitude (for example, where one result is about 50 million for me, his is about 5 million).
We are using the same raster, and the vector files have been projected to match the raster projected coordinate system exactly - to avoid issues from projecting on the fly. The only difference we can find between our protocols is that he is on ArcGis 10.1 and I am on Arc 10.2.1. As far as I know we both have all licenses.
From @Whuber's detailed explanation at Does Zonal Statistics value change when applied on subset of data using ArcGIS Spatal Analyst?, I think it has something to do with changes in the "black box" of zonal statistics calculations.