Oscillations in the optimization procedure of SPM12

[david-hofmann]

The optimization procedure to compute hyper parameter updates seems to be regularly running into periodic orbits. Is that the reason for terminating after 8 steps (and not allowing the user to even change this? It is hard coded)?
The relevant code part can be found in the MATLAB SPM12 file spm_nlsi_GN.m by searching the comment string M-step: Fisher scoring scheme to find h = max{F(p,h)}. In the Julia code it is here: https://github.com/Neuroblox/Dynamic-Causal-Modeling/blob/f1f51405c365f9e598d5ff8fe184a98f396bd5e1/VariationalBayes_for_spectralDCM.jl#L294

Need to have a closer look at this part of the code - maybe there is some fix to this?