no this is so far not implemented. What you could do it get the residuals
out of mne.fit_dipole and reapply the function on residuals. It's not
so trivial though.
You cannot just fit to the residuals! The fields of the previously fitted dipoles have to be projected out of the data and the gain vectors of new candidates. Or, equivalently, the previously fitted dipoles have to be included (as fixed) the new model.
Because of this kind of partly heuristic strategies needed, a GUI is necessary and has to be very responsive and different kinds of feedback needs to be provided.
The MNE-CPP software will eventually include an ?mne_analyze? GUI which will be able to do this.