Through observations in natural rocks, experiments and previous models it is well known that volume increasing hydration reactions, such as serpentinization, lead to fracture nucleation, i.e., reaction-induced fracturing. In their manuscript McElwee et al. bring this process a step forward by investigating how tectonic stresses in various settings influence fracture propagation. Through numerical models they test different stress configurations and find that large fracture networks branching into the surrounding rock form in tensile regimes. To the contrary, in compressional regimes such networks do not form, or only when the reaction is already well advanced, with sever implications on the hydration stage of mid ocean ridges and bending faults. These results are significant and certainly of interest for the community. I only have a few minor comments.

The manuscript is well written and I really enjoyed reading it. Specifically, I acknowledge the detailed discussion on model limitations. All models were run at 1 MPa confining pressure while it is known from experiments that high confining pressures inhibit fracture nucleation. However, I miss a similar discussion on the effect of temperature. We know that the serpentinization rate is sensitive to temperature and maximum reaction rates are reached at 270 – 300 °C. Within the mantle wedge we expect strong temperature gradients, such that reaction rate varies in space as do elastic parameters. In other words, when the reaction is fastest the mechanical behavior may favour visco-elasto-plastic rather than brittle responses to the reaction. At higher temperature, the reaction rate slows down, further supporting non-brittle behavior due to decreased strain rates.

To me it was not clear how the model deals with volume expansion on the scale of individual disks. The mechanical approach explains in detail how the elastic properties change continously from non-reacted to fully reacted disks. The chemical approach explains how fast this transition occurs. However, the serpentinization reaction is strongly volume increasing and hence, the disks are expected to expand. While certain bonds will break and form new fluid pathways, others will ultimately close, which is the often discussed processes of clogging. How exactly is this treated in the model?

Furthermore, the volume change may be slightly dependent on pressure and temperature. Possibly this goes too far for this manuscript, but it might be interesting to test how temperature and pressure will affect the volume change and thus the fracture propagation in various tectonic settings.

Minor comments

Line 10 (and throughout the manuscript): to refer to the process, change “reaction-induced fracture” to “reaction-induced fraturing”.

Line 40: It could be helpful for the reader to have a reference to figure 6 here.

Figure 6: In this figure, the compressional and extensional regimes within the mantle wedge could be labelled/highlighted in order to help the reader.

Line 110: How are the values of P_min and P_max determined?

This manuscript explores the effect of differential stress on reaction hydration dynamics. The results show that the patterns of fractures and fluid pathways depends on the differential stress: 1) in tension (sigma1 < sigma 2 where sigma1 is normal to the side boundaries of the model domain in a pure shear sense), the branching fracture patterns are dominant, which helps with fluid reactions in the interior of the domain, thus further fracturing the interior, and 2) in compression, (sigma1 > sigma 2), spalling patterns are dominant, which limits fluid reactions from happening in the interior of the domain that can further fracture the domain. The paper discusses the implications for a few relevant tectonic settings. Overall, the paper is very good and would be a useful contribution with minor revisions.

Major comments (and questions)

1. Methodology: It would be useful for readers to have a system of equations being solved here so that it is clear how the equations presented in the main text fit into the larger system of equations. The method can be clearer, i.e. the circular disks seem to not overlap but it would be helpful to explicitly state this. What are the boundary effects if there is no layer of unreactive disks and how was the thickness of this layer chosen? Why are the three other boundaries impermeable? It is unclear what the imaginary pipes between disks are (each disk pair has a pipe between them? how is the length of the pipe chosen? is this the full length of the two disk in the direction normal to both disks or perhaps just the overlap?) and are these already scaled to be 1% of the actual pore space (is the aperture is already 1%)? How are these imaginary pipes connected or how do they become connected? I assume these pipes are constant in aperture through its length. How do the water filled cracks become connected? When there is an isolated crack, does it ever increase in volume more than the amount of water that is already there since no fluid is coming in to fill in?
2. Methodology: It is unclear how the volume increase is taken into account for the chemical reactions. Is it an isotropic volume increase (the disk increases in sizes and can overlap)? These models are in two dimensions, are there expected differences if you were to increase to a third dimension (spheres instead of disk)? How does the total solid + fluid volume imposed? Is it some local volume defined by a region surrounding the reacting disk? It is hard to understand how mass is conserved with all these parameterizations. 
3. The disks are randomly placed initially. If these models were to be repeated 1000 times, is there a spread of the model results? In particular, would you get the same answers from every model as shown in Figure 5? 
4. Why is it that the authors chose to run simulations at 1MPa which is not reflective of the environments that they are trying to capture? Why not run simulations at larger confining pressures? 

Minor comments

1.  The figures are good at illustrating what is written in the main text but they can be improved. The figures themselves could be larger along with the labels. It would be clearer if there are labels for the models, i.e. T1,H1,C1 etc or tension/hydrostatic/compression like in figure 4. Typo in Figure 5 legend: there is a missing minus sign for the 5.
2. Line 159: How much faster are reaction rates here compared to nature? References for this? The ratios of fluid flow rates to reaction rates are a nice way to explore this but it is unclear how they fit into the system of equations and how realistic they are as compared to nature. 
3. What is Delta? This is not explained anywhere until stated as `bulk reaction completion' in Figure 5 caption. This needs to be explained much earlier since it appears in Figure 2,3,4. It is also confusing that there is a Delta in the supplementary material that is `average reaction degree' in section S2 and Figure S2. Please fix this. 
4.  A recent paper, Olive et al 2025, showed possible compression at the mid-ocean ridges that would be interesting to discuss. 
5.  It might be useful to the community to have a discussion in relation to what the results mean in the context of the fluid flow modeling at subduction zone that one of the authors is part of [Wilson et al 2014, Cerpa et al 2017,2019, 2025]. 
6. In the supplementary section 1, when the authors talked about fully reacted case, do you mean to 3\% completion or everything is reacted?
7. Do the horizontal dimensions change the model results like the vertical extent does?