diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py
index cbfc978a882..7dced4e26b6 100644
--- a/litellm/proxy/proxy_server.py
+++ b/litellm/proxy/proxy_server.py
@@ -5040,7 +5040,6 @@ class ProxyConfig:
                         # these are litellm callbacks - "langfuse", "sentry", "wandb"
                         else:
                             litellm.logging_callback_manager.add_litellm_success_callback(callback)
-                            _add_custom_logger_callback_to_specific_event(callback, "success")
                             if "prometheus" in callback:
                                 from litellm.integrations.prometheus import (
                                     PrometheusLogger,
@@ -5068,7 +5067,6 @@ class ProxyConfig:
                         # these are litellm callbacks - "langfuse", "sentry", "wandb"
                         else:
                             litellm.logging_callback_manager.add_litellm_failure_callback(callback)
-                            _add_custom_logger_callback_to_specific_event(callback, "failure")
                     print(  # noqa: T201
                         f"{blue_color_code} Initialized Failure Callbacks - {litellm.failure_callback} {reset_color_code}"
                     )
diff --git a/litellm/proxy/utils.py b/litellm/proxy/utils.py
index 86d954c0913..58bb3823e81 100644
--- a/litellm/proxy/utils.py
+++ b/litellm/proxy/utils.py
@@ -173,6 +173,7 @@ from litellm.types.mcp import (
 )
 from litellm.types.proxy.policy_engine.pipeline_types import PipelineExecutionResult
 from litellm.types.utils import LLMResponseTypes, LoggedLiteLLMParams
+from litellm.utils import _add_custom_logger_callback_to_specific_event
 
 if TYPE_CHECKING:
     from mcp.types import CallToolResult
@@ -851,6 +852,16 @@ class ProxyLogging:
                 litellm.logging_callback_manager.add_litellm_async_success_callback(callback)
                 litellm.logging_callback_manager.add_litellm_async_failure_callback(callback)
 
+        # Runs after load_config applied every litellm_settings key: logger __init__s read e.g. s3_callback_params
+        success_callbacks: Final = tuple(cast(list[str | CustomLogger], litellm.success_callback))
+        failure_callbacks: Final = tuple(cast(list[str | CustomLogger], litellm.failure_callback))
+        for callback in success_callbacks:
+            if isinstance(callback, str):
+                _add_custom_logger_callback_to_specific_event(callback, "success")
+        for callback in failure_callbacks:
+            if isinstance(callback, str):
+                _add_custom_logger_callback_to_specific_event(callback, "failure")
+
     async def update_request_status(self, litellm_call_id: str, status: Literal["success", "fail"]):
         # only use this if slack alerting is being used
         if self.alerting is None:
diff --git a/tests/test_litellm/proxy/test_proxy_server.py b/tests/test_litellm/proxy/test_proxy_server.py
index bd784249aa5..ae09d2be1b6 100644
--- a/tests/test_litellm/proxy/test_proxy_server.py
+++ b/tests/test_litellm/proxy/test_proxy_server.py
@@ -2952,26 +2952,44 @@ async def test_custom_ui_sso_sign_in_handler_config_loading():
 
 
 @pytest.mark.asyncio
-async def test_load_config_eagerly_initializes_string_success_callback(tmp_path, monkeypatch):
-    from litellm.integrations.generic_api.generic_api_callback import GenericAPILogger
+async def test_startup_initializes_string_callbacks_after_all_litellm_settings_load(tmp_path, monkeypatch):
+    from litellm.integrations.s3_v2 import S3Logger
+    from litellm.litellm_core_utils import litellm_logging
     from litellm.proxy.proxy_server import ProxyConfig
+    from litellm.proxy.utils import ProxyLogging
 
     config_file = tmp_path / "config.yaml"
     config_file.write_text(
         "model_list: []\n"
         "litellm_settings:\n"
         "  success_callback:\n"
-        "    - generic_api\n"
+        "    - s3_v2\n"
+        "  failure_callback:\n"
+        "    - s3_v2\n"
+        "  s3_callback_params:\n"
+        "    s3_bucket_name: ordering-regression-bucket\n"
+        "    s3_region_name: us-west-2\n"
     )
-    monkeypatch.setenv("GENERIC_LOGGER_ENDPOINT", "http://127.0.0.1:8899/")
 
     monkeypatch.setattr(litellm, "success_callback", [])
     monkeypatch.setattr(litellm, "_async_success_callback", [])
     monkeypatch.setattr(litellm, "failure_callback", [])
     monkeypatch.setattr(litellm, "_async_failure_callback", [])
+    monkeypatch.setattr(litellm, "callbacks", [])
+    monkeypatch.setattr(litellm, "s3_callback_params", None)
+    monkeypatch.setattr(litellm_logging, "_in_memory_loggers", [])
+
     await ProxyConfig().load_config(router=MagicMock(), config_file_path=str(config_file))
-    assert any(isinstance(callback, GenericAPILogger) for callback in litellm._async_success_callback)
-    assert "generic_api" not in litellm.success_callback
+    ProxyLogging(user_api_key_cache=MagicMock())._init_litellm_callbacks(llm_router=None)
+
+    success_loggers = [cb for cb in litellm._async_success_callback if isinstance(cb, S3Logger)]
+    failure_loggers = [cb for cb in litellm._async_failure_callback if isinstance(cb, S3Logger)]
+    assert len(success_loggers) == 1
+    assert len(failure_loggers) == 1
+    assert success_loggers[0].s3_bucket_name == "ordering-regression-bucket"
+    assert success_loggers[0].s3_region_name == "us-west-2"
+    assert "s3_v2" not in litellm.success_callback
+    assert "s3_v2" not in litellm.failure_callback
 
 
 @pytest.mark.asyncio