diff --git a/Lib/test/test_capi/test_mem.py b/Lib/test/test_capi/test_mem.py
index 66ab072733e17bd..e6389b4e0c1e924 100644
--- a/Lib/test/test_capi/test_mem.py
+++ b/Lib/test/test_capi/test_mem.py
@@ -24,9 +24,12 @@ def check(self, code):
             out = assert_python_failure(
                 '-c', code,
                 PYTHONMALLOC=self.PYTHONMALLOC,
-                # FreeBSD: instruct jemalloc to not fill freed() memory
-                # with junk byte 0x5a, see JEMALLOC(3)
+                # Instruct the system allocator to not fill freed() memory
+                # with junk bytes:
+                # FreeBSD: jemalloc, see JEMALLOC(3).
                 MALLOC_CONF="junk:false",
+                # OpenBSD: see MALLOC.CONF(5).
+                MALLOC_OPTIONS="j",
             )
         stderr = out.err
         return stderr.decode('ascii', 'replace')
@@ -102,7 +105,9 @@ def check_pyobject_is_freed(self, func_name):
         assert_python_ok(
             '-c', code,
             PYTHONMALLOC=self.PYTHONMALLOC,
+            # See the comment in check() above.
             MALLOC_CONF="junk:false",
+            MALLOC_OPTIONS="j",
         )
 
     def test_pyobject_null_is_freed(self):
diff --git a/Lib/test/test_concurrent_futures/test_init.py b/Lib/test/test_concurrent_futures/test_init.py
index ca612db17ce8021..c9fedc7213256c7 100644
--- a/Lib/test/test_concurrent_futures/test_init.py
+++ b/Lib/test/test_concurrent_futures/test_init.py
@@ -1,5 +1,6 @@
 import contextlib
 import logging
+import multiprocessing
 import queue
 import time
 import unittest
@@ -147,8 +148,8 @@ def test_spawn(self):
         self._test(ProcessPoolSpawnFailingInitializerTest)
 
     @support.skip_if_sanitizer("TSAN doesn't support threads after fork", thread=True)
-    @unittest.skipIf(sys.platform == "cygwin",
-                     "Forkserver is not available on Cygwin")
+    @unittest.skipUnless("forkserver" in multiprocessing.get_all_start_methods(),
+                         "forkserver start method is not available")
     def test_forkserver(self):
         self._test(ProcessPoolForkserverFailingInitializerTest)
 
diff --git a/Lib/test/test_profiling/test_tracing_profiler.py b/Lib/test/test_profiling/test_tracing_profiler.py
index 8a7070142b60eec..6a4d968f12ef156 100644
--- a/Lib/test/test_profiling/test_tracing_profiler.py
+++ b/Lib/test/test_profiling/test_tracing_profiler.py
@@ -1,4 +1,5 @@
 """Test suite for the cProfile module."""
+import multiprocessing
 import sys
 import unittest
 
@@ -220,8 +221,8 @@ def test_process_spawn_pickle(self):
         # gh-140729: test use Process in cProfile.
         self._test_process_run_pickle('spawn')
 
-    @unittest.skipIf(sys.platform == 'win32',
-                     "No 'forkserver' start method on Windows")
+    @unittest.skipUnless("forkserver" in multiprocessing.get_all_start_methods(),
+                         "forkserver start method is not available")
     def test_process_forkserver_pickle(self):
         # gh-140729: test use Process in cProfile.
         self._test_process_run_pickle('forkserver')
diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2026-07-22-14-30-00.gh-issue-154429.Qk9mZ2.rst b/Misc/NEWS.d/next/Core_and_Builtins/2026-07-22-14-30-00.gh-issue-154429.Qk9mZ2.rst
new file mode 100644
index 000000000000000..04781b03356b9fa
--- /dev/null
+++ b/Misc/NEWS.d/next/Core_and_Builtins/2026-07-22-14-30-00.gh-issue-154429.Qk9mZ2.rst
@@ -0,0 +1,2 @@
+Slightly speed up deallocation of objects that are not tracked by the garbage
+collector, such as :class:`int`, :class:`float` and :class:`str`.
diff --git a/Objects/object.c b/Objects/object.c
index bd23c2e23881949..fadd9273a36607c 100644
--- a/Objects/object.c
+++ b/Objects/object.c
@@ -3292,13 +3292,20 @@ _Py_Dealloc(PyObject *op)
     PyTypeObject *type = Py_TYPE(op);
     unsigned long gc_flag = type->tp_flags & Py_TPFLAGS_HAVE_GC;
     destructor dealloc = type->tp_dealloc;
-    PyThreadState *tstate = _PyThreadState_GET();
-    intptr_t margin = _Py_RecursionLimit_GetMargin(tstate);
-    if (margin < 2 && gc_flag) {
-        _PyTrash_thread_deposit_object(tstate, (PyObject *)op);
-        return;
+    PyThreadState *tstate = NULL;
+    intptr_t margin = 0;
+    if (gc_flag) {
+        tstate = _PyThreadState_GET();
+        margin = _Py_RecursionLimit_GetMargin(tstate);
+        if (margin < 2) {
+            _PyTrash_thread_deposit_object(tstate, (PyObject *)op);
+            return;
+        }
     }
 #ifdef Py_DEBUG
+    if (tstate == NULL) {
+        tstate = _PyThreadState_GET();
+    }
 #if !defined(Py_GIL_DISABLED) && !defined(Py_STACKREF_DEBUG)
     /* This assertion doesn't hold for the free-threading build, as
      * PyStackRef_CLOSE_SPECIALIZED is not implemented */
@@ -3340,7 +3347,7 @@ _Py_Dealloc(PyObject *op)
     Py_XDECREF(old_exc);
     Py_DECREF(type);
 #endif
-    if (tstate->delete_later && margin >= 4 && gc_flag) {
+    if (gc_flag && tstate->delete_later && margin >= 4) {
         _PyTrash_thread_destroy_chain(tstate);
     }
 }