r.grow.distance with r.mask returns errors on a global scale, but not with a smaller extent or without a mask

We are trying to calculate distances to settlements on a global scale (with a 10 arc second resolution), to eliminate 2/3 of the area (water/ocean) to calculate we used land as a mask (r.mask) but this isn’t working on this scale.

• Without a mask it works fine but it takes 8 hours to run

With a mask we get the following errors:

• WARNING: ZSTD compression error -14: Unsupported frame parameter
• ERROR: Error uncompressing raster data for row 33288 of
• CalledModuleError: Module run None [‘r.grow.distance’, ‘–o’, ‘–q’, ‘-m’, ‘distance=r101535547’, ‘metric=geodesic’, ‘input=r101535531’] ended with error

We checked the following:

• Row 33288, there seems to be nothing strange in that row (there is data and all settlement are on land)
• We used smaller datasets (with row 33288) with about 1/4 of the globe and with the mask and they all worked fine
• Running it in ArcGIS with a mask, this worked fine

It runs on a E64s_V3 Azure machine with 432 GiB RAM reading/writing to a shared storage

• Windows Server 2016 Datacenter

• Grass 7.6

• ZSTD 1.3.2.0 (the name of the dll in the OSGEO4W folder)