Add retries for unaccepted worker connections

Author: msimacek

graalpython/com.oracle.graal.python.test/src/runner.py

@@ -727,6 +727,8 @@ def run_in_subprocess_and_watch(self):
             port = server.getsockname()[1]
             assert port
 
+            retries = 3
+
             while self.remaining_test_ids and not self.stop_event.is_set():
                 last_remaining_count = len(self.remaining_test_ids)
                 with open(tmp_dir / 'out', 'w+') as self.out_file:
@@ -744,8 +746,17 @@ def run_in_subprocess_and_watch(self):
                         cmd.append('--failfast')
                     self.process = subprocess.Popen(cmd, stdout=self.out_file, stderr=self.out_file)
 
-                    server.settimeout(600.0)
-                    with server.accept()[0] as sock:
+                    server.settimeout(180.0)
+                    try:
+                        sock = server.accept()[0]
+                    except TimeoutError:
+                        interrupt_process(self.process)
+                        retries -= 1
+                        if retries:
+                            continue
+                        sys.exit("Worker failed to connect to runner multiple times")
+
+                    with sock:
                         conn = Connection(sock)
 
                         conn.send([TestSpecifier(t.test_file, t.test_name) for t in self.remaining_test_ids])