I thank the referee for a third inspection of my paper. I just chose
the same stratification as Gourgouliatos & collaborators, but I was not
anywhere near the limits of the Pencil Code. To demonstrate this, I have
added a Run CZ, which is similar to Run Cz, but with a four times shorter
scale height. The same scalings of phi and xi are reproduced as before,
see Eqs.(36) and (37), but with new prefactors phi0 and xi0 that are
now added in Table 5. I've added a sentences in various places in the
paper (pages 2, 3, 11-13) to fit this new model smoothly into the rest
of the text.

> The new requested figure (Fig. 13) has also been added to the manuscript.
> My last comment about this issue is that the figure shows that the
> employed profile (as in Gourgouliatos et al.) is quite limited and not
> really representing a realistic neutron star. As we can see in the figure,
> the electron density varies only 2 orders of magnitude in the domain,
> and the resistivity a bit more than one order of magnitude. In a neutron
> star crust, density varies 4-5 orders of magnitude, and resistivity 2
> to 5 orders of magnitude, depending on the temperature. Perhaps this
> is the reason of the weak effect of stratification because the profiles
> actually correspond only to the inner crust. If the grid is extended to
> the outer crust, where gradients are much stronger and the Hall parameter
> much larger, the study would be more close to a real neutron star.

I agree. Looking again at Pons & Geppert (2007), I see that the
conductivity varies by about 2.7 orders of magnitude over 0.85 km for
1e8 K and by about 3.7 orders of magnitude for 1e7 K. In my new model,
the conductivity now varies by nearly three orders of magnitude.

> I suppose that the main reason for this choice, instead of using
> a realistic profile, given by a solution of the stellar equilibrium
> equations (which is fairly easy to produce) are the intrinsic numerical
> limitations of the code to work in more extreme cases. I am well aware
> that this was the real reason for the choice of profiles in Gourgouliatos
> et al. papers, where these approximated formulae come from.

The main problem for a Hall cascade simulation is that the typical time
scales become very disparate, so not much happens in the deeper parts
while near the surface there is a lot of change in a short amount of
time. Therefore the total run time must be very long to capture the slow
evolution at the bottom. This is now explained at the end of Sect. 3.
Also, I have now overplotted the stratification for a model with a
four times smaller scale height, so the stratification is now stronger;
see the blue text where I say that n_e varies by 4 orders of magnitude
and eta by nearly 3. The power law for the inverse cascade is still the
same as before, but the coefficients are of course different and those
are now also included in Table 5, as was already mentioned above.

> In any case, the qualitative results are interesting and I can
> recommend the paper for publications after commenting on this caveat
> (the profiles actually represent a limited portion of the crust, just
> the inner crust where gradients are smaller). And I strongly recommend
> the author to keep this in mind for possible future works and start using
> density profiles coming from a solution of a TOV equation, for example,
> as well as realistic resistivity formulae.

Thanks for providing this insider's view into the limitations of the code
used by Gourgouliatos & collaborators. Regarding the calculation of a
realistic stratification I have now added a sentence on this on page 2,
where I refer to the Living Review in Relativity by Chamel & Haensel.