fix(tracing): clean up processors after tracing is disabled

src/agents/tracing/provider.py

@@ -507,10 +507,6 @@ def force_flush(self) -> None:
             log_model_and_tool_action_error(logger, "Error flushing trace provider", e)
 
     def shutdown(self, timeout: float | None = None) -> None:
-        self._refresh_disabled_flag()
-        if self._disabled:
-            return
-
         try:
             _safe_debug("Shutting down trace provider")
             self._multi_processor.shutdown(timeout=timeout)

tests/test_disabled_trace_provider_shutdown.py

@@ -0,0 +1,16 @@
+from __future__ import annotations
+
+from unittest.mock import MagicMock
+
+from agents.tracing.provider import DefaultTraceProvider
+
+
+def test_disabled_trace_provider_still_shuts_down_registered_processors() -> None:
+    provider = DefaultTraceProvider()
+    processor = MagicMock()
+    provider.register_processor(processor)
+    provider.set_disabled(True)
+
+    provider.shutdown()
+
+    processor.shutdown.assert_called_once_with()