When a calculation doesn't reach convergence within the number used in the workchain (80 steps) it does a similar restart.
Repeating the same parameters , this could be up to 4 times.
Perhaps at a second stage (restart) the workchain should just be kill to avoid repeating unsuccessful calculations.
On the other hand, should we consider if the system might need some extra steps? Should the restart try to include 50% more steps and if it doesn't success should it kill the job?