On multithreading jobs the transformed bounds can vary slightly between threads on some datasets. This results in a lot of missing tiles and other issues as different threads can end up processing the same tile, while other tiles get skipped. A mutex around the transform calculation seems sufficient to give the same bounds on each thread.
I'm assuming this is a GDAL issue. gdalinfo --version indicates GDAL 2.0.1
On multithreading jobs the transformed bounds can vary slightly between threads on some datasets. This results in a lot of missing tiles and other issues as different threads can end up processing the same tile, while other tiles get skipped. A mutex around the transform calculation seems sufficient to give the same bounds on each thread.
I'm assuming this is a GDAL issue. gdalinfo --version indicates GDAL 2.0.1