I made two additional jobs, each running after the previous is completed. These two new jobs need to qualify only the nodes that have failed.
So, my primary job runs at midnight, and is set to backup ALL nodes.
Then, my secondary backup job runs at 2am, but instead of ALL nodes, I have it set to all nodes that have status = failed.
Finally, my third backup job runs at 4am, and is also set to run for all nodes that have status = failed.
I find this usually weeds out all of the "bad timings", so to speak, and anything that remains failed most likely has a valid problem.
The important part of the example below, that you will need to do, is the last line. The first two lines are not related to making this work for your requirements.
Someone from the staff would have to confirm, but I am pretty sure this is on their radar to improve with future releases.
I hope this helps.
-Will