Dear Nicolai,
Thank you for the revised version. I am ready to accept it. I noted a few technical corrections listed below. Most could be done when correcting author proofs, but because I also request you get rid of PSU from salinity in the figures, we will need a new technical revision.

Minor technical edits:
in the abstract and in the conclusions refer to weak and strong temperature differences (rather than gradients), as the values are given in degC (not degC/s or /m).

At the end section 2.1 (or at a suitable place in section 2.1) insert "Salinity is measured using the practical salinity scale" and delete PSU as units throughout the text.
Please also remove PSU from the figures (and table 3).

Line 255 and throughout (including Table2): is the significant digit on your response time, \tau_{95} not too ambitious, especially when the range and STD are so large? I would suggest to round them to one significant digit.

Li291, definition of MLD: please clarify as the depth where \sigma_0 increases by 0.01 kg/m3 from its near-surface value (at XX m or vertically average between XX--YY m?)

li290 and 292- inconsistent use of isopycnal 1027.46 and 27.46 (nothing wrong, but better to stick with one. I suggest 27.46)

Li439: poisoned using
----------