Hi Jon,
I'm not positive, but it sounds like you're running into a couple of different but highly related issues here. For the requests marked FAILED in the DQR history, you'd probably need to check the XNAT logs when you're trying to import one of those sessions for us to have a better idea of what's going wrong.
For the "dicom_upload" issue, XNAT requires all sessions in an individual project to have unique session labels. So, if you have multiple sessions all labeled "dicom_upload", archiving all of them will fail, after the first. The DQR plugin is currently rather rigid where it will attempt to use (0020,0010) Study ID as the session label, and if it doesn't find that, it falls back to "dicom_upload". Are these studies all missing a value for Study ID on the PACS side?
Thanks,
Charlie Moore