fix(tests): isolate flaky tests - restore global state in setup/teardown (#21791)

* fix(tests): isolate flaky files endpoint tests from global proxy state

* test(secret_managers): add mocked unit test for write/read JSON secret cycle

* fix(tests): restore litellm.callbacks in TestSpendLogsPayload setup/teardown

* fix(tests): clear app.openapi_schema in TestSwaggerChatCompletions setup/teardown

* fix(tests): add flaky marker to test_async_increment_tokens_with_ttl_preservation

tests/test_litellm/proxy/hooks/test_parallel_request_limiter_v3.py

@@ -1116,6 +1116,7 @@ async def test_dynamic_rate_limiting_v3():
     ), "RPM limit should be enforced when dynamic mode and failures detected"
 
 
+@pytest.mark.flaky(reruns=3)
 @pytest.mark.asyncio
 async def test_async_increment_tokens_with_ttl_preservation():
     """

tests/test_litellm/proxy/spend_tracking/test_spend_management_endpoints.py

@@ -1211,6 +1211,12 @@ async def _wait_for_mock_call(mock, timeout=10, interval=0.1):
 
 
 class TestSpendLogsPayload:
+    def setup_method(self):
+        self._original_callbacks = litellm.callbacks[:]
+
+    def teardown_method(self):
+        litellm.callbacks = self._original_callbacks
+
     @pytest.mark.asyncio
     async def test_spend_logs_payload_e2e(self):
         litellm.callbacks = [_ProxyDBLogger(message_logging=False)]

tests/test_litellm/proxy/test_swagger_chat_completions.py

@@ -17,6 +17,12 @@ from litellm.proxy.proxy_server import app
 class TestSwaggerChatCompletions:
     """Test suite for validating /chat/completions schema in Swagger documentation."""
 
+    def setup_method(self):
+        app.openapi_schema = None
+
+    def teardown_method(self):
+        app.openapi_schema = None
+
     @pytest.fixture
     def client(self):
         """FastAPI test client for the proxy server."""
@@ -315,7 +321,8 @@ class TestSwaggerChatCompletions:
         This ensures Swagger UI works correctly with reverse proxies and subpath deployments.
         """
         from unittest.mock import patch
-        from litellm.proxy.proxy_server import get_openapi_schema, custom_openapi, app
+
+        from litellm.proxy.proxy_server import app, custom_openapi, get_openapi_schema
 
         # Test cases: (server_root_path, expected_servers_url)
         # Note: empty string is falsy in Python, so servers won't be set