diff --git a/ducktape/tests/runner.py b/ducktape/tests/runner.py
index 992ab24d..f931bd61 100644
--- a/ducktape/tests/runner.py
+++ b/ducktape/tests/runner.py
@@ -291,8 +291,21 @@ def run_all_tests(self):
                                 if proc.is_alive():
                                     try:
                                         os.kill(proc.pid, signal.SIGUSR1)
-                                    except OSError:
-                                        pass
+                                    except OSError as e:
+                                        self._log(logging.ERROR,
+                                                  "Failed to send SIGUSR1 to pid %d: %s" %
+                                                  (proc.pid, e))
+                                        continue
+                                else:
+                                    exitcode = proc.exitcode
+                                    if exitcode is not None and exitcode < 0:
+                                        reason = "killed by %s" % signal.Signals(-exitcode).name
+                                    else:
+                                        reason = "exitcode=%s" % exitcode
+                                    self._log(logging.ERROR,
+                                              "Worker pid %d already dead (%s)" %
+                                              (proc.pid, reason))
+                                    break
                                 time.sleep(0.005)
 
                         # All processes are on the same machine, so treat communication failure as a fatal error
diff --git a/tests/runner/check_runner.py b/tests/runner/check_runner.py
index 4912de66..cfbd6e8a 100644
--- a/tests/runner/check_runner.py
+++ b/tests/runner/check_runner.py
@@ -287,6 +287,28 @@ def check_runner_timeout_requests_stack_dumps(self, capfd):
         captured = capfd.readouterr()
         assert "Current thread" in captured.err
 
+    @pytest.mark.parametrize('crash,expected_reason', [
+        (lambda: os._exit(1), "exitcode=1"),
+        (lambda: os.kill(os.getpid(), signal.SIGKILL), "killed by SIGKILL"),
+    ])
+    def check_runner_timeout_dead_worker(self, crash, expected_reason):
+        """Check that dead workers are reported when requesting stack dumps."""
+        mock_cluster = LocalhostCluster(num_nodes=1000)
+        session_context = tests.ducktape_mock.session_context(max_parallel=1000, test_runner_timeout=1)
+
+        test_methods = [TestThingy.test_delayed]
+        ctx_list = self._do_expand(test_file=TEST_THINGY_FILE, test_class=TestThingy, test_methods=test_methods,
+                                   cluster=mock_cluster, session_context=session_context)
+        mock_logger = Mock()
+        runner = TestRunner(mock_cluster, session_context, mock_logger, ctx_list, 1)
+
+        with pytest.raises(TimeoutError), \
+                patch('ducktape.tests.runner.run_client', lambda *a, **kw: crash()):
+            runner.run_all_tests()
+
+        log_messages = [str(c) for c in mock_logger.log.call_args_list]
+        assert any("already dead" in msg and expected_reason in msg for msg in log_messages)
+
     @pytest.mark.parametrize('fail_greedy_tests', [True, False])
     def check_fail_greedy_tests(self, fail_greedy_tests):
         mock_cluster = LocalhostCluster(num_nodes=1000)