If an image scan fails due to clair pod crashing, the image is marked as failed and never rescheduled

There is a logic bug in this section of code:

https://github.com/quay/quay/blob/master/data/secscan_model/secscan_v4_model.py#L293-L313

If clair crashes during the the scan, the image is marked as failed in the database and is never rescheduled for scan, again, requiring the quay admin to update the database manually to force the rescan.

rhn-support-dshumake may have more details/insight.