Following Landsat’s practice, Sentinel-2 multispectral satellite products are delivered as raster images projected onto the Universal Transversal Mercator (UTM) spatial reference system, which divides Earth into 60 longitudinal zones. Locally, this guarantees high spatial accuracy, while also easing the interoperability with many regional and governmental datums. On top, the Sentinel-2 product grid uses the Military Grid Reference System (MGRS) tiling scheme to facilitate manageable data slices and straightforward multitemporal image stacking. Although most convenient for small-area applications, activities with a larger geographic scope suffer from this approach and its overhead, as both data duplication and ambiguity appear along UTM zone overlaps and MGRS tile borders. In practice, such areas that are covered by multiple and incongruent grid pixels are known but just tolerated, and their degree has not been measured so far.In this paper, we illuminate the nature and patterns of these overlaps, and calculate the resulting spatial redundancy over the global land surface. We found that the total land area is enlarged in the Sentinel-2 grid definition by 33%, which is a value similar to the simple and single-zoned Plate Carrée projection. The number of co-located grid pixels for a single location ranges from 1 up to 6, with on average more redundancy at mid- and high-latitudes. With regard to global satellite archives in times of big data and increased energy costs, the examined grid appears as a suboptimal choice, inducing complexity and overhead at an unreasonable level. Owing to the grid design, e.g., the yearly Sentinel-2 user product volume (Level-1C and -2A) is inflated by 1 petabyte, entailing cascading downstream costs of storage, bandwidth, and computing.