diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 09bf95ddc5b..25818ea2192 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -290,7 +290,7 @@ jobs:
 
       - name: Install sktime and dependencies
         shell: bash
-        run: python -m pip install .[all_extras_pandas2,dev,dl] --no-cache-dir
+        run: python -m pip install .[all_extras_pandas2,dev,dl,compatibility_tests] --no-cache-dir
 
       - name: Show dependencies
         run: python -m pip list
diff --git a/README.md b/README.md
index 858c4caaec9..09fa39b62e7 100644
--- a/README.md
+++ b/README.md
@@ -4,7 +4,7 @@
 
 > A unified interface for machine learning with time series
 
-:rocket: **Version 0.31.1 out now!** [Check out the release notes here](https://www.sktime.net/en/latest/changelog.html).
+:rocket: **Version 0.31.2 out now!** [Check out the release notes here](https://www.sktime.net/en/latest/changelog.html).
 
 sktime is a library for time series analysis in Python. It provides a unified interface for multiple time series learning tasks. Currently, this includes time series classification, regression, clustering, annotation, and forecasting. It comes with [time series algorithms](https://www.sktime.net/en/stable/estimator_overview.html) and [scikit-learn] compatible tools to build, tune and validate time series models.
 
diff --git a/docs/source/changelog.rst b/docs/source/changelog.rst
index 2f10451bc05..fe725d48eb3 100644
--- a/docs/source/changelog.rst
+++ b/docs/source/changelog.rst
@@ -19,6 +19,20 @@ For upcoming changes and next releases, see our `milestones <https://github.com/
 For our long-term plan, see our :ref:`roadmap`.
 
 
+Version 0.31.2 - 2024-08-13
+---------------------------
+
+Hotfix release for using ``make_reduction`` with not fully ``sklearn`` compliant
+tabular regressors such as from ``catboost``.
+
+For last non-maintenance content updates, see 0.31.1.
+
+Contents
+~~~~~~~~
+
+* [BUG] fix ``make_reduction`` type inference for non-sklearn estimators
+
+
 Version 0.31.1 - 2024-08-10
 ---------------------------
 
diff --git a/pyproject.toml b/pyproject.toml
index ee49a827498..74647d6511b 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "sktime"
-version = "0.31.1"
+version = "0.31.2"
 description = "A unified framework for machine learning with time series"
 readme = "README.md"
 keywords = [
@@ -305,6 +305,9 @@ numpy1 = [
 pandas1 = [
   "pandas<2.0.0",
 ]
+compatibility_tests = [
+  "catboost",
+]
 
 [project.urls]
 "API Reference" = "https://www.sktime.net/en/stable/api_reference.html"
diff --git a/sktime/__init__.py b/sktime/__init__.py
index 5a6b8c9d188..25a96fc3311 100644
--- a/sktime/__init__.py
+++ b/sktime/__init__.py
@@ -1,6 +1,6 @@
 """sktime."""
 
-__version__ = "0.31.1"
+__version__ = "0.31.2"
 
 __all__ = ["show_versions"]
 
diff --git a/sktime/forecasting/compose/_reduce.py b/sktime/forecasting/compose/_reduce.py
index 26befa94522..6b261b0ee89 100644
--- a/sktime/forecasting/compose/_reduce.py
+++ b/sktime/forecasting/compose/_reduce.py
@@ -1631,7 +1631,7 @@ def _infer_scitype(estimator):
     if is_sklearn_estimator(estimator):
         return f"tabular-{sklearn_scitype(estimator)}"
     else:
-        inferred_skt_scitype = scitype(estimator)
+        inferred_skt_scitype = scitype(estimator, raise_on_unknown=False)
         if inferred_skt_scitype in ["object", "estimator"]:
             return "tabular-regressor"
         if inferred_skt_scitype == "regressor":
@@ -2260,7 +2260,21 @@ def get_test_params(cls, parameter_set="default"):
             "windows_identical": False,
         }
         params5 = {"estimator": est, "window_length": 0}
-        return [params1, params2, params3, params4, params5]
+
+        params = [params1, params2, params3, params4, params5]
+
+        # this fails because catboost is not sklearn compatible
+        # and fails set_params contracts already in sklearn;
+        # so it also fails them in sktime...
+        # left here for future reference, e.g., test for non-compliant estimators
+        #
+        # if _check_soft_dependencies("catboost", severity="none"):
+        #     from catboost import CatBoostRegressor
+        #
+        #     est = CatBoostRegressor(learning_rate=1, depth=6, loss_function="RMSE")
+        #     params6 = {"estimator": est, "window_length": 3}
+        #     params.append(params6)
+        return params
 
 
 class RecursiveReductionForecaster(BaseForecaster, _ReducerMixin):
diff --git a/sktime/forecasting/compose/tests/test_reduce.py b/sktime/forecasting/compose/tests/test_reduce.py
index 9ec1565397a..7a042c85e78 100644
--- a/sktime/forecasting/compose/tests/test_reduce.py
+++ b/sktime/forecasting/compose/tests/test_reduce.py
@@ -441,6 +441,33 @@ def test_make_reduction_infer_scitype_for_sklearn_pipeline():
     assert forecaster._estimator_scitype == "tabular-regressor"
 
 
+@pytest.mark.skipif(
+    not run_test_module_changed(["sktime.forecasting", "sktime.split"])
+    or not _check_soft_dependencies("catboost", severity="none"),
+    reason="run test only if forecasting or split module has changed",
+)
+def test_make_reduction_with_catboost():
+    """Test make_reduction with catboost.
+
+    catboost is an example of a package that does not fully comply with the
+    sklearn API. We therefore need to rely on the branch of scitype inference
+    that assumes the estimator is a tabular regressor.
+    """
+    from catboost import CatBoostRegressor
+
+    estimator = CatBoostRegressor(
+        learning_rate=1, depth=6, loss_function="RMSE", verbose=False
+    )
+
+    forecaster = make_reduction(estimator, scitype="infer")
+    assert forecaster._estimator_scitype == "tabular-regressor"
+
+    fh = [1, 2, 3]
+    y, X = make_forecasting_problem(make_X=True)
+    y_train, y_test, X_train, X_test = temporal_train_test_split(y, X, fh=fh)
+    forecaster.fit(y_train, X_train, fh=fh).predict(fh, X_test)
+
+
 @pytest.mark.skipif(
     not run_test_module_changed(["sktime.forecasting.compose._reduce"]),
     reason="run test only if reduce module has changed",