Fermi energy and restart

1. QE version: latest development version (30 June 2021)
2. Local workstation, ifort (IFORT) 19.1.3.304 20200925
3. Input and output data (please see below)
4. Description of the problem: I study NiO at the DFT level of theory (PBEsol, USPP), k-mesh: 10x10x10, parallelization over plane waves only (using 8 cores). I compare the direct calculation and the one with the restart. All is the same (total energy, KS eigenvalues, etc.) but the Fermi energy: 11.6317 eV (direct) VS 11.5486 eV. results_direct.tarresults_with_restart.tar Any ideas why the Fermi energy is different? This seems to be a bug.