[Pw_forum] problem with scf in ruthenium - textbook example

[Alexander Kulesza]

Dear users and developers,

sorry for bothering you with the following problem on a text-book example.
I try to obtain atomization energies for metals and use the input for 
Ruthenium as attached at the end of the mail. My problem is the weird 
non-convergence behaviour for the hexagonal system. Here the scf gets 
numerically unstable (for US PP) or a "negative dr2" error (PAW) 
appears. Interestingly, for a hypothetical bcc Ru the problem does not 
appear and correctly predicts a way-too-low atomization energy.

My question: Am I missing some obvious and stupid error in my input ? 
are the two errors related with each other and/or the pseudopotential ?
For the PAW case i have read that commenting one line out may help to 
solve aborting upon "negative dr2" but i do not see how this could 
overcome the numerical instability of the US-PP scf.

Your help is greatly appreciated.
With kind regards
Alexander

&control
     calculation='scf'
[...]
  /
  &system
     ibrav=4, celldm(1)=5.1003, celldm(3)=1.584
     nat=2, ntyp=1,
     nspin = 2, starting_magnetization(1)=0.0
     ecutwfc = 40.0, ecutrho = 400.0, nbnd = 24,
     occupations='smearing', smearing='methfessel-paxton', degauss=0.01
  /
  &electrons
     startingwfc='random'
     diagonalization="david"
     conv_thr = 1.0e-8
     mixing_beta = 0.8, mixing_mode= 'plain'
/
ATOMIC_SPECIES
Ru 58.69 Ru.pbe-spn-rrkjus_psl.0.3.0.UPF
ATOMIC_POSITIONS {crystal}
Ru 0.3333333333  0.6666666667  0.25
Ru 0.6666666667  0.3333333333  0.75
K_POINTS {automatic}
10 10 10 0 0 0

I use the recent compilation of pseudopotentials provided by A. Dal 
Corso pslibrary.