diff --git a/ament_tools/verbs/test_pkg/cli.py b/ament_tools/verbs/test_pkg/cli.py
index ac8ecb1..188f3f9 100644
--- a/ament_tools/verbs/test_pkg/cli.py
+++ b/ament_tools/verbs/test_pkg/cli.py
@@ -47,6 +47,12 @@ def prepare_arguments(parser, args, skip_build_pkg_arguments=False):
         type=int, default=0, metavar='N',
         help='Rerun failing tests up to N times',
     )
+    parser.add_argument(
+        '--ignore-return-codes',
+        action='store_true',
+        default=False,
+        help='Ignore return codes from testing packages',
+    )
     return parser
 
 
@@ -75,18 +81,21 @@ def main(opts):
             return
         try:
             handle_build_action(on_test_ret, context)
-        except SystemExit:
+        except SystemExit as e:
             # check if tests should be rerun
             if opts.retest_until_pass > context.test_iteration:
                 context.test_iteration += 1
                 print("+++ Testing '%s' again (retry #%d of %d)" %
                       (pkg_name, context.test_iteration, opts.retest_until_pass))
                 continue
-            # there is no way to distinguish why the test returned non zero
-            # the test invocation itself could have failed:
-            # return e.code
-            # but it could have also run successful and only failed some tests:
-            return
+            # Automated systems can use --ignore-return-codes to allow them to react to
+            # a failure to *run* a test but not a failure generated by a test that ran as
+            # intended. Otherwise, we'll combine the two cases to help users to notice
+            # when anything went wrong during a test run.
+            if opts.ignore_return_codes:
+                return
+            else:
+                return e.code
         # check if tests should be rerun
         if opts.retest_until_fail > context.test_iteration:
             context.test_iteration += 1