Yes it happens when lapw0 starts after couple of cycles.
Probably your are using an old version of WIEN2k with the old scheme
to find the solution of the nonlinear equation in brj.f. You should get
the new version brj.f which avoids such problems.
Should I re-run with a clean_lapw or just delete a job and then re-run scf
It should not matter as long as your calculation continues and
Wien mailing list
SEARCH the MAILING-LIST at: