It could be useful to store the JobID that initially created a file in the filesystem, as well as the objects that are written to. This information is easily available on the MDS and OSS at create/write time, and allows users to find out after the fact what Job created the file, which allows them to link it back to parameters that were used at the time.

I'd propose to save this in the lustre.job xattr, since this can be visible to regular users, unlike the "trusted" xattrs.