When producing a parquet file containing a Boolean column the files will produce slightly different checksums each time if the row count per row group is quite high. I traced this down to the ...