I have several TB data in my DAG, and file checksumming happens in both DAG building and jobs completion stage, causing the DAG building and job completion during running be extremely slow (more than 30mins or even 1h). See the diff in https://github.com/laf070810/snakemake for the affected parts.

I suggest adding some file size checking to avoid large file checksumming, or adding a config option to completely disable file checksumming.

Related to #2920