(automatic) k-point convergence test?
Since we now 'quasi' allow several k-point sets in the inp.xml, maybe it would be nice to directly generate a second slightly denser mesh (purpose k-con-test) with inpgen and calculate again with this k-point set after the charge convergence (if some switch is set, as in the band structure case)? These results might be used to estimate if the quantities one is interested in are converged with respect to the k-mesh. Also Fleur could issue a warning that if the differences in the total energy per atom is to large. Is this with the current implementation easy to do?
Drawback: optimal parallelization choice becomes harder to impossible. Of course in an aiida workflow we can do this convergence check with using several Fleur jobs, the last charge density would be copied in this case (or from scratch), leading also to more overhead in the database if we do it for every SCF we run. If one could do it within one job it would be nicer, but of course more work on the programming side.