diff --git a/metaflow/metaflow_config.py b/metaflow/metaflow_config.py
index 066e5659216..385dfa98f5f 100644
--- a/metaflow/metaflow_config.py
+++ b/metaflow/metaflow_config.py
@@ -540,6 +540,7 @@
     "userconf",
     "conda",
     "package",
+    "card",
 ]
 
 for typ in DEBUG_OPTIONS:
diff --git a/metaflow/plugins/cards/card_creator.py b/metaflow/plugins/cards/card_creator.py
index 597c25096de..ec5f5389141 100644
--- a/metaflow/plugins/cards/card_creator.py
+++ b/metaflow/plugins/cards/card_creator.py
@@ -5,6 +5,7 @@
 import sys
 import os
 from metaflow import current
+from metaflow.debug import debug
 from typing import Callable, Tuple, Dict
 
 
@@ -87,6 +88,10 @@ def create(
             component_strings = current.card._serialize_components(card_uuid)
             # Since the mode is a render, we can check if we need to write to the metadata store.
             save_metadata, metadata_dict = self._should_save_metadata(card_uuid)
+        debug.card_exec(
+            "create card_uuid=%s mode=%s final=%s sync=%s runtime_card=%s save_metadata=%s"
+            % (card_uuid, mode, final, sync, runtime_card, save_metadata)
+        )
         data = current.card._get_latest_data(card_uuid, final=final, mode=mode)
         runspec = "/".join([current.run_id, current.step_name, current.task_id])
         self._run_cards_subprocess(
@@ -179,6 +184,7 @@ def _run_cards_subprocess(
         if save_metadata:
             cmd += ["--save-metadata", json.dumps(metadata_dict)]
 
+        debug.card_exec(cmd)
         response, fail = self._run_command(
             cmd,
             card_uuid,
@@ -219,9 +225,28 @@ def _run_command(self, cmd, card_uuid, env, wait=True, timeout=None):
             except subprocess.CalledProcessError as e:
                 rep = e.output
                 fail = True
+                if debug.card and rep is not None:
+                    debug.card_exec(
+                        "render subprocess for %s %s output:\n%s"
+                        % (
+                            card_uuid,
+                            "FAILED" if fail else "succeeded",
+                            rep.decode("utf-8"),
+                        )
+                    )
             except subprocess.TimeoutExpired as e:
                 rep = e.output
                 fail = True
+                if debug.card and rep is not None:
+                    debug.card_exec(
+                        "render subprocess for %s %s output:\n%s"
+                        % (
+                            card_uuid,
+                            "FAILED" if fail else "succeeded",
+                            rep.decode("utf-8"),
+                        )
+                    )
+
             return rep, fail
         else:
             _async_proc, _async_started = CardProcessManager._get_card_process(
@@ -238,8 +263,13 @@ def _run_command(self, cmd, card_uuid, env, wait=True, timeout=None):
                 else:
                     # silently refuse to run an async process if a previous one is still running
                     # and timeout hasn't been reached
+                    debug.card_exec(
+                        "skipping async render for %s: previous process still running"
+                        % card_uuid
+                    )
                     return "".encode(), False
             else:
+                debug.card_exec("launching async render for %s" % card_uuid)
                 CardProcessManager._register_card_process(
                     card_uuid,
                     subprocess.Popen(
diff --git a/metaflow/plugins/cards/card_decorator.py b/metaflow/plugins/cards/card_decorator.py
index daa667fa2a7..9c2ec1ff4b1 100644
--- a/metaflow/plugins/cards/card_decorator.py
+++ b/metaflow/plugins/cards/card_decorator.py
@@ -10,6 +10,7 @@
 from metaflow.user_configs.config_options import ConfigInput
 from metaflow.user_configs.config_parameters import dump_config_values
 from metaflow.util import to_unicode
+from metaflow.debug import debug
 
 from .component_serializer import CardComponentCollector, get_card_class
 from .card_creator import CardCreator
@@ -241,6 +242,17 @@ def task_pre_step(
                 self._is_editable = True
             self._is_runtime_card = card_class.RUNTIME_UPDATABLE
 
+        debug.card_exec(
+            "task_pre_step step=%s type=%s found=%s editable=%s runtime=%s"
+            % (
+                step_name,
+                card_type,
+                card_class is not None,
+                self._is_editable,
+                self._is_runtime_card,
+            )
+        )
+
         # We have a step counter to ensure that on calling the final card decorator's `task_pre_step`
         # we call a `finalize` function in the `CardComponentCollector`.
         # This can help ensure the behaviour of the `current.card` object is according to specification.
@@ -322,6 +334,16 @@ def task_finished(
             card_options=self.card_options,
             logger=self._logger,
         )
+        debug.card_exec(
+            "task_finished step=%s card_uuid=%s type=%s is_task_ok=%s -> %s"
+            % (
+                step_name,
+                self._card_uuid,
+                self.attributes["type"],
+                is_task_ok,
+                "rendering" if is_task_ok else "skipping (task failed)",
+            )
+        )
         if is_task_ok:
             self.card_creator.create(mode="render", final=True, **create_options)
             self.card_creator.create(mode="refresh", final=True, **create_options)