diff --git a/drf_spectacular/plumbing.py b/drf_spectacular/plumbing.py
index 4e3464af..639c4f7b 100644
--- a/drf_spectacular/plumbing.py
+++ b/drf_spectacular/plumbing.py
@@ -905,13 +905,20 @@ def deep_import_string(string: str) -> Any:
 
 
 def load_enum_name_overrides():
-    return _load_enum_name_overrides(get_language())
+    # id() of the current ENUM_NAME_OVERRIDES dict is used (in addition to language) as part
+    # of the cache key, so a patched settings object (e.g. via SpectacularAPIView's
+    # custom_settings, serving multiple schemas with different overrides) gets its own cache
+    # entry instead of reusing whatever was cached for a different settings object. The dict
+    # itself can't be used directly as a cache key/argument since it (or its values, which may
+    # be raw choice lists) is not hashable.
+    return _load_enum_name_overrides(get_language(), id(spectacular_settings.ENUM_NAME_OVERRIDES))
 
 
 @functools.lru_cache()
-def _load_enum_name_overrides(language: str):
+def _load_enum_name_overrides(language: str, enum_name_overrides_id: int):
+    enum_name_overrides = spectacular_settings.ENUM_NAME_OVERRIDES
     overrides = {}
-    for name, choices in spectacular_settings.ENUM_NAME_OVERRIDES.items():
+    for name, choices in enum_name_overrides.items():
         if isinstance(choices, str):
             choices = deep_import_string(choices)
         if not choices:
@@ -944,7 +951,7 @@ def _load_enum_name_overrides(language: str):
         ]
         overrides[list_hash(hashable_values)] = name
 
-    if len(spectacular_settings.ENUM_NAME_OVERRIDES) != len(overrides):
+    if len(enum_name_overrides) != len(overrides):
         error(
             'ENUM_NAME_OVERRIDES has duplication issues. Encountered multiple names '
             'for the same choice set. Enum naming might be unexpected.'
diff --git a/tests/test_postprocessing.py b/tests/test_postprocessing.py
index eb84dc98..5fe5059d 100644
--- a/tests/test_postprocessing.py
+++ b/tests/test_postprocessing.py
@@ -157,6 +157,29 @@ class XView(generics.RetrieveAPIView):
     assert len(schema['components']['schemas']) == 2
 
 
+def test_global_enum_naming_override_not_stale_across_settings_objects(no_warnings, clear_caches):
+    """
+    Regression test for #1244. load_enum_name_overrides() used to be cached only on
+    language, so loading overrides for one ENUM_NAME_OVERRIDES dict would poison the
+    cache for a *different* ENUM_NAME_OVERRIDES dict with the same language - exactly
+    what happens serving multiple schemas with per-schema custom_settings, since
+    SpectacularAPIView's custom_settings context manager doesn't clear this cache.
+    """
+    with mock.patch(
+        'drf_spectacular.settings.spectacular_settings.ENUM_NAME_OVERRIDES',
+        {'LanguageEnum': 'tests.test_postprocessing.language_choices'},
+    ):
+        overrides_a = load_enum_name_overrides()
+        assert overrides_a == {list_hash(list(language_choices)): 'LanguageEnum'}
+
+    with mock.patch(
+        'drf_spectacular.settings.spectacular_settings.ENUM_NAME_OVERRIDES',
+        {'VoteEnum': 'tests.test_postprocessing.vote_choices'},
+    ):
+        overrides_b = load_enum_name_overrides()
+        assert overrides_b == {list_hash(list(vote_choices)): 'VoteEnum'}
+
+
 @mock.patch('drf_spectacular.settings.spectacular_settings.ENUM_NAME_OVERRIDES', {
     'LanguageEnum': 'tests.test_postprocessing.blank_null_language_choices'
 })