In a simple setup (hexagonal Co monolayer), where noncollinear calculations are done and the magnetization is along y-direction (i.e. non-collinear alpha and beta-angles are pi/2), the 1st iteration yields an unreasonably high charge distance:

Iteration:           1  Distance:   102.670344519425

Additionally, the magnetic moment has an opposite sign as compared to the input-value.

[zimmer]$ grep moment out.xml_wrong
            <magneticMoment atomType="     1" moment="  -1.3548983100" spinUpCharge="   3.2537215250" spinDownCharge="   4.6086198350"/>

With an older Fleur version, these issues did not arise.

Iteration:           1  Distance:   16.0322487705896

[zimmer]$ grep moment out.xml_correct
            <magneticMoment atomType="     1" moment="   1.7577425000" spinUpCharge="   4.8507903337" spinDownCharge="   3.0930478337"/>

inp.xmlsym.outout.xml_correctout.xml_wrong