In picard/formats/_init_.py#48 we open the file for reading, and then read the first 128 bytes in order to try to determine the internal format from the first 128 bytes.
However on the file function we do not specify a buffer size, which means the default buffer size is used. The default buffer size can be determined with:
and on my Win64 system is 8192.
It would be better to use the following code, which will avoid reading more sectors of disk than minimum needed (depending on o/s, disk format and cluster size) and minimises the amount of data transferred across the network: