diff --git a/automation/scripts/adaptability_check.py b/automation/scripts/adaptability_check.py
index d843782f8d..1714342405 100644
--- a/automation/scripts/adaptability_check.py
+++ b/automation/scripts/adaptability_check.py
@@ -1207,7 +1207,7 @@ def _comments(source: str) -> list[tuple[int, int, str]]:
         for token in tokenize.generate_tokens(io.StringIO(source).readline):
             if token.type == tokenize.COMMENT:
                 out.append((token.start[0], token.start[1], token.string))
-    except (tokenize.TokenError, IndentationError, SyntaxError):
+    except tokenize.TokenError, IndentationError, SyntaxError:
         for number, line in enumerate(source.splitlines(), start=1):
             stripped = line.lstrip()
             if stripped.startswith("#"):
diff --git a/changelog.d/executor-stuck-test.md b/changelog.d/executor-stuck-test.md
new file mode 100644
index 0000000000..fecd6023a1
--- /dev/null
+++ b/changelog.d/executor-stuck-test.md
@@ -0,0 +1,9 @@
+### Fixed
+
+- **The stuck-launcher executor test no longer fails on a slow CI runner.** The fake
+  launcher execs the harness, so the process the test keeps alive carried the job's
+  one-second `RLIMIT_CPU`; when the harness start spent that second, the kernel killed
+  the "stuck" launcher inside the grace window and `stuck` read 0. The test now gives
+  the run the host's full CPU cap, which a real `sandbox do` outside the sandbox always
+  has, and asserts that limit on the survivor, so it fails every time without the fix
+  (#12260).
diff --git a/core/adaptability.py b/core/adaptability.py
index f4895fcaff..ec34cd3ab0 100644
--- a/core/adaptability.py
+++ b/core/adaptability.py
@@ -876,7 +876,7 @@ def _comments(source: str) -> list[tuple[int, int, str]]:
         for token in tokenize.generate_tokens(io.StringIO(source).readline):
             if token.type == tokenize.COMMENT:
                 out.append((token.start[0], token.start[1], token.string))
-    except (tokenize.TokenError, IndentationError, SyntaxError):
+    except tokenize.TokenError, IndentationError, SyntaxError:
         for number, line in enumerate(source.splitlines(), start=1):
             stripped = line.lstrip()
             if stripped.startswith("#"):
diff --git a/tests/unit/agents/renderer/test_executor.py b/tests/unit/agents/renderer/test_executor.py
index 4c54742259..24b88792ed 100644
--- a/tests/unit/agents/renderer/test_executor.py
+++ b/tests/unit/agents/renderer/test_executor.py
@@ -7,6 +7,7 @@
 import asyncio
 import base64
 import os
+import resource
 import signal
 import stat
 import time
@@ -148,13 +149,20 @@ async def test_a_launcher_that_outlives_its_kill_blocks_renders_until_it_exits(
         executor = SandboxExecutor(replace(config, kill_grace_s=0.2))
         monkeypatch.setattr(executor_module, "READER_GRACE_S", 0.2)
         monkeypatch.setattr(executor_module.os, "killpg", lambda pid, sig: None)  # the kill has no effect
+        # The fake launcher execs the harness, so the survivor carries the run's RLIMIT_CPU, which a real
+        # `sandbox do` on the host never does. With the one second of a one-second job, the kernel's SIGKILL
+        # ended the "stuck" launcher inside the grace window whenever the harness start (matplotlib, a fresh
+        # font cache) spent that second, as it did on a loaded CI runner (2026-10-23). The full host cap
+        # leaves the survivor nothing but the test's own kill to die of.
+        monkeypatch.setattr(executor_module, "run_cpu_seconds", lambda limit_s, timeout_s: limit_s)
 
         run = await executor.run_theme(job(SLEEPER, timeout_s=1.0), "light")
 
         assert run.timed_out and executor.stuck == 1
+        survivor = executor._stuck[0]
+        assert resource.prlimit(survivor.pid, resource.RLIMIT_CPU) == (config.rlimit_cpu_s,) * 2
         with pytest.raises(Unavailable, match="stuck"):
             await executor.run_theme(job(PLOT), "light")
-        survivor = executor._stuck[0]
         os.kill(survivor.pid, signal.SIGKILL)
         await survivor.wait()
         assert executor.stuck == 0