automatic maintenance scheduling caused overlap of updates on both nodes, not giving enough time for the second one to automatically block updates because of failure on the first one => adjustment of timing formula discussed.
it broke at a time when admin availability was low.