diff --git a/.buildkite/data.rayci.yml b/.buildkite/data.rayci.yml
index ec9469e834b0..ff1137139659 100644
--- a/.buildkite/data.rayci.yml
+++ b/.buildkite/data.rayci.yml
@@ -28,30 +28,6 @@ steps:
     wanda: ci/docker/datatfxbsl.build.wanda.yaml
 
   # tests
-  - label: ":database: data: arrow v9 tests"
-    tags:
-      - data
-    instance_type: medium
-    parallelism: 2
-    commands:
-      - bazel run //ci/ray_ci:test_in_docker -- //python/ray/data/... //python/ray/air/... data
-        --workers "$${BUILDKITE_PARALLEL_JOB_COUNT}"
-        --worker-id "$${BUILDKITE_PARALLEL_JOB}" --parallelism-per-worker 3
-        --build-name data9build
-        --except-tags data_integration,doctest,data_non_parallel
-    depends_on: data9build
-
-  - label: ":database: data: arrow v9 tests (data_non_parallel)"
-    tags:
-      - data
-      - data_non_parallel
-    instance_type: medium
-    commands:
-      - bazel run //ci/ray_ci:test_in_docker -- //python/ray/data/... //python/ray/air/... data
-        --build-name data9build
-        --only-tags data_non_parallel
-    depends_on: data9build
-
   - label: ":database: data: arrow v19 tests"
     tags:
       - python
diff --git a/python/ray/_private/arrow_utils.py b/python/ray/_private/arrow_utils.py
index 0c75ef6c4fa7..cd6fb8a07363 100644
--- a/python/ray/_private/arrow_utils.py
+++ b/python/ray/_private/arrow_utils.py
@@ -82,10 +82,6 @@ def add_creatable_buckets_param_if_s3_uri(uri: str) -> str:
         URI is an S3 URL; uri will be returned unchanged otherwise.
     """
 
-    pyarrow_version = get_pyarrow_version()
-    if pyarrow_version is not None and pyarrow_version < parse_version("9.0.0"):
-        # This bucket creation query parameter is not required for pyarrow < 9.0.0.
-        return uri
     parsed_uri = urlparse(uri)
     if parsed_uri.scheme == "s3":
         uri = _add_url_query_params(uri, {"allow_bucket_creation": True})
diff --git a/python/ray/air/tests/test_air_usage.py b/python/ray/air/tests/test_air_usage.py
index a2e14f3607c2..aca2830a0271 100644
--- a/python/ray/air/tests/test_air_usage.py
+++ b/python/ray/air/tests/test_air_usage.py
@@ -87,11 +87,6 @@ def test_tag_storage_type(storage_path_filesystem_expected, mock_record, monkeyp
 
     storage_path, storage_filesystem, expected = storage_path_filesystem_expected
 
-    if Version(pyarrow.__version__) < Version("9.0.0") and storage_path.startswith(
-        "gs://"
-    ):
-        pytest.skip("GCS support requires pyarrow >= 9.0.0")
-
     storage = StorageContext(
         storage_path=storage_path,
         experiment_dir_name="test",
diff --git a/python/ray/air/tests/test_object_extension.py b/python/ray/air/tests/test_object_extension.py
index b95f4c44a958..a8d439790be2 100644
--- a/python/ray/air/tests/test_object_extension.py
+++ b/python/ray/air/tests/test_object_extension.py
@@ -8,14 +8,10 @@
 from ray.air.util.object_extensions.arrow import (
     ArrowPythonObjectArray,
     ArrowPythonObjectType,
-    _object_extension_type_allowed,
 )
 from ray.air.util.object_extensions.pandas import PythonObjectArray
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension not supported."
-)
 def test_object_array_validation():
     # Test unknown input type raises TypeError.
     with pytest.raises(TypeError):
@@ -25,9 +21,6 @@ def test_object_array_validation():
     PythonObjectArray([object(), object()])
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension not supported."
-)
 def test_arrow_scalar_object_array_roundtrip():
     arr = np.array(
         ["test", 20, False, {"some": "value"}, None, np.zeros((10, 10))], dtype=object
@@ -41,9 +34,6 @@ def test_arrow_scalar_object_array_roundtrip():
     assert np.all(out[-1] == arr[-1])
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension not supported."
-)
 def test_arrow_python_object_array_slice():
     arr = np.array(["test", 20, "test2", 40, "test3", 60], dtype=object)
     ata = ArrowPythonObjectArray.from_objects(arr)
@@ -51,9 +41,6 @@ def test_arrow_python_object_array_slice():
     assert ata[2:4].to_pylist() == ["test2", 40]
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension not supported."
-)
 def test_arrow_pandas_roundtrip():
     obj = types.SimpleNamespace(a=1, b="test")
     t1 = pa.table({"a": ArrowPythonObjectArray.from_objects([obj, obj]), "b": [0, 1]})
@@ -61,18 +48,12 @@ def test_arrow_pandas_roundtrip():
     assert t1.equals(t2)
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension not supported."
-)
 def test_pandas_python_object_isna():
     arr = np.array([1, np.nan, 3, 4, 5, np.nan, 7, 8, 9], dtype=object)
     ta = PythonObjectArray(arr)
     np.testing.assert_array_equal(ta.isna(), pd.isna(arr))
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension not supported."
-)
 def test_pandas_python_object_take():
     arr = np.array([1, 2, 3, 4, 5], dtype=object)
     ta = PythonObjectArray(arr)
@@ -85,9 +66,6 @@ def test_pandas_python_object_take():
     )
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension not supported."
-)
 def test_pandas_python_object_concat():
     arr1 = np.array([1, 2, 3, 4, 5], dtype=object)
     arr2 = np.array([6, 7, 8, 9, 10], dtype=object)
diff --git a/python/ray/air/tests/test_tensor_extension.py b/python/ray/air/tests/test_tensor_extension.py
index fb5b6bbd43ab..d5f0767f611a 100644
--- a/python/ray/air/tests/test_tensor_extension.py
+++ b/python/ray/air/tests/test_tensor_extension.py
@@ -517,20 +517,8 @@ def test_arrow_tensor_array_getitem(chunked, restore_data_context, tensor_format
     if chunked:
         t_arr = pa.chunked_array(t_arr)
 
-    pyarrow_version = get_pyarrow_version()
-    if (
-        chunked
-        and pyarrow_version >= parse_version("8.0.0")
-        and pyarrow_version < parse_version("9.0.0")
-    ):
-        for idx in range(outer_dim):
-            item = t_arr[idx]
-            assert isinstance(item, pa.ExtensionScalar)
-            item = item.type._extension_scalar_to_ndarray(item)
-            np.testing.assert_array_equal(item, arr[idx])
-    else:
-        for idx in range(outer_dim):
-            np.testing.assert_array_equal(t_arr[idx], arr[idx])
+    for idx in range(outer_dim):
+        np.testing.assert_array_equal(t_arr[idx], arr[idx])
 
     # Test __iter__.
     for t_subarr, subarr in zip(t_arr, arr):
@@ -554,19 +542,8 @@ def test_arrow_tensor_array_getitem(chunked, restore_data_context, tensor_format
 
     np.testing.assert_array_equal(t_arr2_npy, arr[1:])
 
-    if (
-        chunked
-        and pyarrow_version >= parse_version("8.0.0")
-        and pyarrow_version < parse_version("9.0.0")
-    ):
-        for idx in range(1, outer_dim):
-            item = t_arr2[idx - 1]
-            assert isinstance(item, pa.ExtensionScalar)
-            item = item.type._extension_scalar_to_ndarray(item)
-            np.testing.assert_array_equal(item, arr[idx])
-    else:
-        for idx in range(1, outer_dim):
-            np.testing.assert_array_equal(t_arr2[idx - 1], arr[idx])
+    for idx in range(1, outer_dim):
+        np.testing.assert_array_equal(t_arr2[idx - 1], arr[idx])
 
 
 @pytest.mark.parametrize("tensor_format", ["v1", "v2"])
@@ -589,20 +566,8 @@ def test_arrow_variable_shaped_tensor_array_getitem(
     if chunked:
         t_arr = pa.chunked_array(t_arr)
 
-    pyarrow_version = get_pyarrow_version()
-    if (
-        chunked
-        and pyarrow_version >= parse_version("8.0.0")
-        and pyarrow_version < parse_version("9.0.0")
-    ):
-        for idx in range(outer_dim):
-            item = t_arr[idx]
-            assert isinstance(item, pa.ExtensionScalar)
-            item = item.type._extension_scalar_to_ndarray(item)
-            np.testing.assert_array_equal(item, arr[idx])
-    else:
-        for idx in range(outer_dim):
-            np.testing.assert_array_equal(t_arr[idx], arr[idx])
+    for idx in range(outer_dim):
+        np.testing.assert_array_equal(t_arr[idx], arr[idx])
 
     # Test __iter__.
     for t_subarr, subarr in zip(t_arr, arr):
@@ -628,19 +593,8 @@ def test_arrow_variable_shaped_tensor_array_getitem(
     for t_subarr, subarr in zip(t_arr2_npy, arr[1:]):
         np.testing.assert_array_equal(t_subarr, subarr)
 
-    if (
-        chunked
-        and pyarrow_version >= parse_version("8.0.0")
-        and pyarrow_version < parse_version("9.0.0")
-    ):
-        for idx in range(1, outer_dim):
-            item = t_arr2[idx - 1]
-            assert isinstance(item, pa.ExtensionScalar)
-            item = item.type._extension_scalar_to_ndarray(item)
-            np.testing.assert_array_equal(item, arr[idx])
-    else:
-        for idx in range(1, outer_dim):
-            np.testing.assert_array_equal(t_arr2[idx - 1], arr[idx])
+    for idx in range(1, outer_dim):
+        np.testing.assert_array_equal(t_arr2[idx - 1], arr[idx])
 
 
 @pytest.mark.parametrize("tensor_format", ["v1", "v2"])
diff --git a/python/ray/air/util/object_extensions/arrow.py b/python/ray/air/util/object_extensions/arrow.py
index 180fcfc96367..fc6ae16d9f54 100644
--- a/python/ray/air/util/object_extensions/arrow.py
+++ b/python/ray/air/util/object_extensions/arrow.py
@@ -10,17 +10,6 @@
 from ray._private.arrow_utils import get_pyarrow_version
 from ray.util.annotations import PublicAPI
 
-MIN_PYARROW_VERSION_SCALAR_SUBCLASS = parse_version("9.0.0")
-
-PYARROW_VERSION = get_pyarrow_version()
-
-
-def _object_extension_type_allowed() -> bool:
-    return (
-        PYARROW_VERSION is not None
-        and PYARROW_VERSION >= MIN_PYARROW_VERSION_SCALAR_SUBCLASS
-    )
-
 
 # Please see https://arrow.apache.org/docs/python/extending_types.html for more info
 @PublicAPI(stability="alpha")
@@ -89,7 +78,7 @@ class ArrowPythonObjectArray(pa.ExtensionArray):
     """Array class for ArrowPythonObjectType"""
 
     def from_objects(
-        objects: typing.Union[np.ndarray, typing.Iterable[typing.Any]]
+        objects: typing.Union[np.ndarray, typing.Iterable[typing.Any]],
     ) -> "ArrowPythonObjectArray":
         if isinstance(objects, np.ndarray):
             objects = objects.tolist()
diff --git a/python/ray/air/util/tensor_extensions/arrow.py b/python/ray/air/util/tensor_extensions/arrow.py
index 146968364424..06afde890bf4 100644
--- a/python/ray/air/util/tensor_extensions/arrow.py
+++ b/python/ray/air/util/tensor_extensions/arrow.py
@@ -132,40 +132,29 @@ def convert_to_pyarrow_array(
 
     except ArrowConversionError as ace:
         from ray.data import DataContext
-        from ray.data.extensions.object_extension import (
-            ArrowPythonObjectArray,
-            _object_extension_type_allowed,
-        )
+        from ray.data.extensions.object_extension import ArrowPythonObjectArray
 
         enable_fallback_config: Optional[
             bool
         ] = DataContext.get_current().enable_fallback_to_arrow_object_ext_type
 
-        if not _object_extension_type_allowed():
-            object_ext_type_fallback_allowed = False
+        # NOTE: By default setting is unset which (for compatibility reasons)
+        #       is allowing the fallback
+        object_ext_type_fallback_allowed = (
+            enable_fallback_config is None or enable_fallback_config
+        )
+
+        if object_ext_type_fallback_allowed:
             object_ext_type_detail = (
-                "skipping fallback to serialize as pickled python"
-                f" objects (due to unsupported Arrow version {PYARROW_VERSION}, "
-                f"min required version is {MIN_PYARROW_VERSION_SCALAR_SUBCLASS})"
+                "falling back to serialize as pickled python objects"
             )
         else:
-            # NOTE: By default setting is unset which (for compatibility reasons)
-            #       is allowing the fallback
-            object_ext_type_fallback_allowed = (
-                enable_fallback_config is None or enable_fallback_config
+            object_ext_type_detail = (
+                "skipping fallback to serialize as pickled python objects "
+                "(due to DataContext.enable_fallback_to_arrow_object_ext_type "
+                "= False)"
             )
 
-            if object_ext_type_fallback_allowed:
-                object_ext_type_detail = (
-                    "falling back to serialize as pickled python objects"
-                )
-            else:
-                object_ext_type_detail = (
-                    "skipping fallback to serialize as pickled python objects "
-                    "(due to DataContext.enable_fallback_to_arrow_object_ext_type "
-                    "= False)"
-                )
-
         # To avoid logging following warning for every block it's
         # only going to be logged in following cases
         #   - It's being logged for the first time, and
@@ -293,7 +282,7 @@ def _coerce_np_datetime_to_pa_timestamp_precision(
 
 
 def _infer_pyarrow_type(
-    column_values: Union[List[Any], np.ndarray]
+    column_values: Union[List[Any], np.ndarray],
 ) -> Optional[pa.DataType]:
     """Infers target Pyarrow `DataType` based on the provided
     columnar values.
@@ -375,7 +364,7 @@ def _len_gt_overflow_threshold(obj: Any) -> bool:
 
 
 def _try_infer_pa_timestamp_type(
-    column_values: Union[List[Any], np.ndarray]
+    column_values: Union[List[Any], np.ndarray],
 ) -> Optional[pa.DataType]:
     if isinstance(column_values, list) and len(column_values) > 0:
         # In case provided column values is a list of elements, this
diff --git a/python/ray/data/_internal/arrow_block.py b/python/ray/data/_internal/arrow_block.py
index d954e7146926..4145c04ee77d 100644
--- a/python/ray/data/_internal/arrow_block.py
+++ b/python/ray/data/_internal/arrow_block.py
@@ -195,21 +195,11 @@ def _build_tensor_row(
         from packaging.version import parse as parse_version
 
         element = row[col_name][0]
-        # TODO(Clark): Reduce this to np.asarray(element) once we only support Arrow
-        # 9.0.0+.
-        pyarrow_version = get_pyarrow_version()
-        if pyarrow_version is None or pyarrow_version >= parse_version("8.0.0"):
-            assert isinstance(element, pyarrow.ExtensionScalar)
-            if pyarrow_version is None or pyarrow_version >= parse_version("9.0.0"):
-                # For Arrow 9.0.0+, accessing an element in a chunked tensor array
-                # produces an ArrowTensorScalar, which we convert to an ndarray using
-                # .as_py().
-                element = element.as_py()
-            else:
-                # For Arrow 8.*, accessing an element in a chunked tensor array produces
-                # an ExtensionScalar, which we convert to an ndarray using our custom
-                # method.
-                element = element.type._extension_scalar_to_ndarray(element)
+        assert isinstance(element, pyarrow.ExtensionScalar)
+        # For Arrow 9.0.0+, accessing an element in a chunked tensor array
+        # produces an ArrowTensorScalar, which we convert to an ndarray using
+        # .as_py().
+        element = element.as_py()
         # For Arrow < 8.0.0, accessing an element in a chunked tensor array produces an
         # ndarray, which we return directly.
         assert isinstance(element, np.ndarray), type(element)
diff --git a/python/ray/data/extensions/__init__.py b/python/ray/data/extensions/__init__.py
index 517b4fe7a3a2..0e8954988e1d 100644
--- a/python/ray/data/extensions/__init__.py
+++ b/python/ray/data/extensions/__init__.py
@@ -8,7 +8,6 @@
     ArrowPythonObjectType,
     PythonObjectArray,
     PythonObjectDtype,
-    _object_extension_type_allowed,
 )
 from ray.data.extensions.tensor_extension import (
     ArrowConversionError,
@@ -40,6 +39,5 @@
     "ArrowPythonObjectScalar",
     "PythonObjectArray",
     "PythonObjectDtype",
-    "_object_extension_type_allowed",
     "get_arrow_extension_tensor_types",
 ]
diff --git a/python/ray/data/extensions/object_extension.py b/python/ray/data/extensions/object_extension.py
index 42ab20a231c6..cd97bb398c34 100644
--- a/python/ray/data/extensions/object_extension.py
+++ b/python/ray/data/extensions/object_extension.py
@@ -2,7 +2,6 @@
     ArrowPythonObjectArray,
     ArrowPythonObjectScalar,
     ArrowPythonObjectType,
-    _object_extension_type_allowed,
 )
 from ray.air.util.object_extensions.pandas import (  # noqa: F401
     PythonObjectArray,
diff --git a/python/ray/data/tests/conftest.py b/python/ray/data/tests/conftest.py
index 905c3e210435..1d0baceb25df 100644
--- a/python/ray/data/tests/conftest.py
+++ b/python/ray/data/tests/conftest.py
@@ -127,9 +127,8 @@ def _s3_fs(aws_credentials, s3_server, s3_path):
 
     kwargs = aws_credentials.copy()
 
-    if get_pyarrow_version() >= parse_version("9.0.0"):
-        kwargs["allow_bucket_creation"] = True
-        kwargs["allow_bucket_deletion"] = True
+    kwargs["allow_bucket_creation"] = True
+    kwargs["allow_bucket_deletion"] = True
 
     fs = pa.fs.S3FileSystem(
         region="us-west-2",
diff --git a/python/ray/data/tests/test_all_to_all.py b/python/ray/data/tests/test_all_to_all.py
index c82d5dc21457..407d958c04fc 100644
--- a/python/ray/data/tests/test_all_to_all.py
+++ b/python/ray/data/tests/test_all_to_all.py
@@ -1881,9 +1881,6 @@ def func(df, a, b, c):
     assert sorted([x["value"] for x in ds.take()]) == [6, 8, 10, 12]
 
 
-_NEED_UNWRAP_ARROW_SCALAR = get_pyarrow_version() <= parse_version("9.0.0")
-
-
 @pytest.mark.parametrize("num_parts", [1, 30])
 @pytest.mark.parametrize("ds_format", ["pyarrow", "pandas", "numpy"])
 def test_groupby_map_groups_multicolumn(
@@ -1902,8 +1899,6 @@ def test_groupby_map_groups_multicolumn(
         num_parts
     )
 
-    should_unwrap_pa_scalars = ds_format == "pyarrow" and _NEED_UNWRAP_ARROW_SCALAR
-
     def _map_group(df):
         # NOTE: Since we're grouping by A and B, these columns will be bearing
         #       the same values.
@@ -1911,8 +1906,8 @@ def _map_group(df):
         b = df["B"][0]
         return {
             # NOTE: PA 9.0 requires explicit unwrapping into Python objects
-            "A": [a.as_py() if should_unwrap_pa_scalars else a],
-            "B": [b.as_py() if should_unwrap_pa_scalars else b],
+            "A": [a],
+            "B": [b],
             "count": [len(df["A"])],
         }
 
@@ -1956,8 +1951,6 @@ def test_groupby_map_groups_multicolumn_with_nan(
         ]
     ).repartition(num_parts)
 
-    should_unwrap_pa_scalars = ds_format == "pyarrow" and _NEED_UNWRAP_ARROW_SCALAR
-
     def _map_group(df):
         # NOTE: Since we're grouping by A and B, these columns will be bearing
         #       the same values
@@ -1965,8 +1958,8 @@ def _map_group(df):
         b = df["B"][0]
         return {
             # NOTE: PA 9.0 requires explicit unwrapping into Python objects
-            "A": [a.as_py() if should_unwrap_pa_scalars else a],
-            "B": [b.as_py() if should_unwrap_pa_scalars else b],
+            "A": [a],
+            "B": [b],
             "count": [len(df["A"])],
         }
 
diff --git a/python/ray/data/tests/test_arrow_block.py b/python/ray/data/tests/test_arrow_block.py
index 5ca8009910cc..787941e8b9a3 100644
--- a/python/ray/data/tests/test_arrow_block.py
+++ b/python/ray/data/tests/test_arrow_block.py
@@ -25,7 +25,6 @@
 from ray.data._internal.arrow_ops.transform_pyarrow import combine_chunked_array
 from ray.data._internal.util import GiB, MiB
 from ray.data.block import BlockAccessor
-from ray.data.extensions.object_extension import _object_extension_type_allowed
 
 
 def simple_array():
@@ -460,9 +459,6 @@ def test_register_arrow_types(tmp_path):
     run_string_as_driver(driver_script)
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension type not supported."
-)
 def test_dict_doesnt_fallback_to_pandas_block(ray_start_regular_shared):
     # If the UDF returns a column with dict, previously, we would
     # fall back to pandas, because we couldn't convert it to
diff --git a/python/ray/data/tests/test_arrow_serialization.py b/python/ray/data/tests/test_arrow_serialization.py
index eb1da5d93dca..4f95ea00d124 100644
--- a/python/ray/data/tests/test_arrow_serialization.py
+++ b/python/ray/data/tests/test_arrow_serialization.py
@@ -26,7 +26,6 @@
 from ray._private.arrow_utils import get_pyarrow_version
 from ray.data.extensions.object_extension import (
     ArrowPythonObjectArray,
-    _object_extension_type_allowed,
 )
 from ray.data.extensions.tensor_extension import (
     ArrowTensorArray,
@@ -415,14 +414,10 @@ def pickled_objects_array():
     (lazy_fixture("list_of_empty_struct_array"), 0.1),
     # Complex nested array
     (lazy_fixture("complex_nested_array"), 0.1),
+    # Array of pickled objects
+    (lazy_fixture("pickled_objects_array"), 0.1),
 ]
 
-if _object_extension_type_allowed():
-    pytest_custom_serialization_arrays.append(
-        # Array of pickled objects
-        (lazy_fixture("pickled_objects_array"), 0.1),
-    )
-
 
 @pytest.mark.parametrize("data,cap_mult", pytest_custom_serialization_arrays)
 def test_custom_arrow_data_serializer(ray_start_regular_shared, data, cap_mult):
@@ -543,9 +538,6 @@ def fn(batch: list):
     assert res == [{"id": 1}], res
 
 
-@pytest.mark.skipif(
-    not _object_extension_type_allowed(), reason="Object extension not supported."
-)
 def test_arrow_object_and_array_support(ray_start_regular_shared):
     obj = types.SimpleNamespace(some_attribute="test")
 
diff --git a/python/ray/data/tests/test_pandas_block.py b/python/ray/data/tests/test_pandas_block.py
index 26b2be7e1e75..0bee6108baa7 100644
--- a/python/ray/data/tests/test_pandas_block.py
+++ b/python/ray/data/tests/test_pandas_block.py
@@ -15,7 +15,6 @@
     PandasBlockColumnAccessor,
 )
 from ray.data._internal.util import is_null
-from ray.data.extensions.object_extension import _object_extension_type_allowed
 
 # Set seed for the test for size as it related to sampling
 np.random.seed(42)
@@ -201,9 +200,6 @@ def test_pandas_block_timestamp_ns(ray_start_regular_shared):
         ), "Timestamp mismatch in PandasBlockBuilder output"
 
 
-@pytest.mark.skipif(
-    _object_extension_type_allowed(), reason="Objects can be put into Arrow"
-)
 def test_dict_fallback_to_pandas_block(ray_start_regular_shared):
     # If the UDF returns a column with dict, this throws
     # an error during block construction because we cannot cast dicts
diff --git a/python/ray/data/tests/test_strict_mode.py b/python/ray/data/tests/test_strict_mode.py
index cd7045daccb6..4c4f7d41e755 100644
--- a/python/ray/data/tests/test_strict_mode.py
+++ b/python/ray/data/tests/test_strict_mode.py
@@ -185,7 +185,6 @@ def test_strict_schema(ray_start_regular_shared):
     from ray.data._internal.pandas_block import PandasBlockSchema
     from ray.data.extensions.object_extension import (
         ArrowPythonObjectType,
-        _object_extension_type_allowed,
     )
     from ray.data.extensions.tensor_extension import ArrowTensorType
 
@@ -203,21 +202,13 @@ def test_strict_schema(ray_start_regular_shared):
 
     ds = ray.data.from_items([{"x": 2, "y": object(), "z": [1, 2]}])
     schema = ds.schema()
-    if _object_extension_type_allowed():
-        assert isinstance(schema.base_schema, pa.lib.Schema)
-        assert schema.names == ["x", "y", "z"]
-        assert schema.types == [
-            pa.int64(),
-            ArrowPythonObjectType(),
-            pa.list_(pa.int64()),
-        ]
-    else:
-        assert schema.names == ["x", "y", "z"]
-        assert schema.types == [
-            pa.int64(),
-            object,
-            object,
-        ]
+    assert isinstance(schema.base_schema, pa.lib.Schema)
+    assert schema.names == ["x", "y", "z"]
+    assert schema.types == [
+        pa.int64(),
+        ArrowPythonObjectType(),
+        pa.list_(pa.int64()),
+    ]
 
     ds = ray.data.from_numpy(np.ones((100, 10)))
     schema = ds.schema()
diff --git a/python/ray/data/tests/test_transform_pyarrow.py b/python/ray/data/tests/test_transform_pyarrow.py
index 9c81a9cf141a..7d26de8c4b16 100644
--- a/python/ray/data/tests/test_transform_pyarrow.py
+++ b/python/ray/data/tests/test_transform_pyarrow.py
@@ -28,7 +28,6 @@
     ArrowTensorArray,
     ArrowTensorType,
     ArrowVariableShapedTensorType,
-    _object_extension_type_allowed,
 )
 
 
@@ -999,9 +998,6 @@ def map_batches(x):
     return ds
 
 
-@pytest.mark.skipif(
-    _object_extension_type_allowed(), reason="Arrow table supports pickled objects"
-)
 @pytest.mark.parametrize(
     "op, data",
     [
diff --git a/python/ray/tests/test_storage.py b/python/ray/tests/test_storage.py
index 565144056fd9..9482abd168f1 100644
--- a/python/ray/tests/test_storage.py
+++ b/python/ray/tests/test_storage.py
@@ -258,32 +258,26 @@ def test_connecting_to_cluster(shutdown_only, storage_type):
 
 
 def test_add_creatable_buckets_param_if_s3_uri():
-    if get_pyarrow_version() >= parse_version("9.0.0"):
-        # Test that the allow_bucket_creation=true query arg is added to an S3 URI.
-        uri = "s3://bucket/foo"
-        assert (
-            add_creatable_buckets_param_if_s3_uri(uri)
-            == "s3://bucket/foo?allow_bucket_creation=true"
-        )
-
-        # Test that query args are merged (i.e. existing query args aren't dropped).
-        uri = "s3://bucket/foo?bar=baz"
-        assert (
-            add_creatable_buckets_param_if_s3_uri(uri)
-            == "s3://bucket/foo?allow_bucket_creation=true&bar=baz"
-        )
-
-        # Test that existing allow_bucket_creation=false query arg isn't overridden.
-        uri = "s3://bucket/foo?allow_bucket_creation=false"
-        assert (
-            add_creatable_buckets_param_if_s3_uri(uri)
-            == "s3://bucket/foo?allow_bucket_creation=false"
-        )
-    else:
-        # Test that the allow_bucket_creation=true query arg is not added to an S3 URI,
-        # since we're using Arrow < 9.
-        uri = "s3://bucket/foo"
-        assert add_creatable_buckets_param_if_s3_uri(uri) == uri
+    # Test that the allow_bucket_creation=true query arg is added to an S3 URI.
+    uri = "s3://bucket/foo"
+    assert (
+        add_creatable_buckets_param_if_s3_uri(uri)
+        == "s3://bucket/foo?allow_bucket_creation=true"
+    )
+
+    # Test that query args are merged (i.e. existing query args aren't dropped).
+    uri = "s3://bucket/foo?bar=baz"
+    assert (
+        add_creatable_buckets_param_if_s3_uri(uri)
+        == "s3://bucket/foo?allow_bucket_creation=true&bar=baz"
+    )
+
+    # Test that existing allow_bucket_creation=false query arg isn't overridden.
+    uri = "s3://bucket/foo?allow_bucket_creation=false"
+    assert (
+        add_creatable_buckets_param_if_s3_uri(uri)
+        == "s3://bucket/foo?allow_bucket_creation=false"
+    )
 
     # Test that non-S3 URI is unchanged.
     uri = "gcs://bucket/foo"
diff --git a/release/ray_release/byod/byod_train_persistence_test.sh b/release/ray_release/byod/byod_train_persistence_test.sh
index c81d4f9f03a6..d72ed36d7b91 100755
--- a/release/ray_release/byod/byod_train_persistence_test.sh
+++ b/release/ray_release/byod/byod_train_persistence_test.sh
@@ -4,4 +4,4 @@
 
 set -exo pipefail
 
-pip3 install -U torch fsspec s3fs gcsfs pyarrow>=9.0.0 pytest
+pip3 install -U torch fsspec s3fs gcsfs pyarrow>=17.0.0 pytest