OK, this is a separate issue. We have tracked this down to a GPU out-of-memory issue related to MiewID. Restarting the MiewID deep learning server solves the issue, restarts the jobs, and the links above are now completed with results.
We are working with our machine learning engineers to understand whether this is a memory leak or whether we simply need a more powerful GPU.
This has been resolved. The Whiskerbook server was allowing up to three simultaneous MiewID/detector jobs (1 fast lane, two slow lane). We limited this to max 2 (1 slow, 1 fast) a week ago and have not observed any job problems of this nature since then.