diff --git a/dev/changelog.html b/dev/changelog.html
index f0602a027..cfb57f4a8 100644
--- a/dev/changelog.html
+++ b/dev/changelog.html
@@ -305,6 +305,7 @@ <h3 class="anchored" data-anchor-id="fixed">Fixed</h3>
 <li>Mutating BMI model control functions (<code>update</code>, <code>update_until</code> and <code>finalize</code>) and extended mutating BMI functions (<code>load_state</code> and <code>save_state</code>) should return <code>nothing</code>.</li>
 <li>Added downloading of testdata to Dockerfile, to ensure an image was able to build.</li>
 <li>The reservoir (<code>reservoir_index_f</code>) and lake (<code>lake_index_f</code>) indices as part of <code>network.river</code> were not correct. These were mapped to their own index in the <code>SimpleReservoir</code> and <code>Lake</code> struct, and not to the corresponding river index. This resulted in incorrect surface water abstractions from reservoir and lake volumes, and surface water abstractions were set at zero at the wrong river locations.</li>
+<li>Wflow ZMQ server: allow JSON reading and writing of <code>NaN</code> and <code>Inf</code> values to avoid a JSON spec error. For example, during initialization of a wflow model, some (diagnostic) model variables are initialized with <code>NaN</code> values.</li>
 </ul>
 </section>
 <section id="changed" class="level3">
diff --git a/dev/search.json b/dev/search.json
index f92e8b577..bfeb272a8 100644
--- a/dev/search.json
+++ b/dev/search.json
@@ -796,7 +796,7 @@
     "href": "changelog.html#unreleased",
     "title": "Changelog",
     "section": "Unreleased",
-    "text": "Unreleased\n\nFixed\n\nInitialization of LateralSSF variables ssf and ssfmax with vertical hydraulic conductivity profile exponential_constant. Removed parameter khfrac from the computation, as it is already part of parameter kh_0.\nMutating BMI model control functions (update, update_until and finalize) and extended mutating BMI functions (load_state and save_state) should return nothing.\nAdded downloading of testdata to Dockerfile, to ensure an image was able to build.\nThe reservoir (reservoir_index_f) and lake (lake_index_f) indices as part of network.river were not correct. These were mapped to their own index in the SimpleReservoir and Lake struct, and not to the corresponding river index. This resulted in incorrect surface water abstractions from reservoir and lake volumes, and surface water abstractions were set at zero at the wrong river locations.\n\n\n\nChanged\n\nRemoved vertical concepts HBV and FLEXTopo.\nRemoved metadata macros exchange and grid_type. The macro grid_type is not required because this functionality is already part of BMI. The macro exchange is replaced by a function used by BMI. Remaining metadata macros get_units and grid_loc are only used by BMI.\nRefactor the vertical SBM concept: divide the long struct SBM into different model components for interception, snow, glacier, (open water) runoff, soil, water demand and allocation stored in the struct LandHydrologySBM. Additionally, the atmospheric forcing and a shared vegetation parameterset are stored as separate fields in struct LandHydrologySBM (with soil model SbmSoilModel). The model component structs have model variables, parameters and boundary_conditions (if applicable), including associated functions for initializing and updating these model components. The original long update function of the SBM soil part has been split into separate functions.\nRefactor the lateral (routing) components: as for the vertical SBM concept split the structs into variables, parameters and boundary_conditions (if applicable).\nRefactor the sediment concept: similar to the refactor of SBM concept, the sediment model has been split into three main structures. SoilLoss was split into rainfall_erosion, overland_flow_erosion and (total) soil_erosion concepts. OverlandFlowSediment has been split into transport_capacity, sediment_flux and to_river concepts. RiverSediment has been split into transport_capacity, potential_erosion, sediment_flux and concentrations concepts.\nTimestepping method parameters for solving the kinematic wave and local inertial approaches for river and overland flow are moved to a TimeStepping struct. The timestepping implementation for the kinematic wave is now similar to the local inertial method: a stable timestep is computed for each sub timestep (or a fixed sub timestep is used) as part of a while loop (for each model timestep).\n\n\n\nAdded\n\nSupport direct output of snow and glacier melt, and add computation of snow water equivalent (SWE).",
+    "text": "Unreleased\n\nFixed\n\nInitialization of LateralSSF variables ssf and ssfmax with vertical hydraulic conductivity profile exponential_constant. Removed parameter khfrac from the computation, as it is already part of parameter kh_0.\nMutating BMI model control functions (update, update_until and finalize) and extended mutating BMI functions (load_state and save_state) should return nothing.\nAdded downloading of testdata to Dockerfile, to ensure an image was able to build.\nThe reservoir (reservoir_index_f) and lake (lake_index_f) indices as part of network.river were not correct. These were mapped to their own index in the SimpleReservoir and Lake struct, and not to the corresponding river index. This resulted in incorrect surface water abstractions from reservoir and lake volumes, and surface water abstractions were set at zero at the wrong river locations.\nWflow ZMQ server: allow JSON reading and writing of NaN and Inf values to avoid a JSON spec error. For example, during initialization of a wflow model, some (diagnostic) model variables are initialized with NaN values.\n\n\n\nChanged\n\nRemoved vertical concepts HBV and FLEXTopo.\nRemoved metadata macros exchange and grid_type. The macro grid_type is not required because this functionality is already part of BMI. The macro exchange is replaced by a function used by BMI. Remaining metadata macros get_units and grid_loc are only used by BMI.\nRefactor the vertical SBM concept: divide the long struct SBM into different model components for interception, snow, glacier, (open water) runoff, soil, water demand and allocation stored in the struct LandHydrologySBM. Additionally, the atmospheric forcing and a shared vegetation parameterset are stored as separate fields in struct LandHydrologySBM (with soil model SbmSoilModel). The model component structs have model variables, parameters and boundary_conditions (if applicable), including associated functions for initializing and updating these model components. The original long update function of the SBM soil part has been split into separate functions.\nRefactor the lateral (routing) components: as for the vertical SBM concept split the structs into variables, parameters and boundary_conditions (if applicable).\nRefactor the sediment concept: similar to the refactor of SBM concept, the sediment model has been split into three main structures. SoilLoss was split into rainfall_erosion, overland_flow_erosion and (total) soil_erosion concepts. OverlandFlowSediment has been split into transport_capacity, sediment_flux and to_river concepts. RiverSediment has been split into transport_capacity, potential_erosion, sediment_flux and concentrations concepts.\nTimestepping method parameters for solving the kinematic wave and local inertial approaches for river and overland flow are moved to a TimeStepping struct. The timestepping implementation for the kinematic wave is now similar to the local inertial method: a stable timestep is computed for each sub timestep (or a fixed sub timestep is used) as part of a while loop (for each model timestep).\n\n\n\nAdded\n\nSupport direct output of snow and glacier melt, and add computation of snow water equivalent (SWE).",
     "crumbs": [
       "Home",
       "Changelog"