diff --git a/source/common/thread_local/thread_local_impl.cc b/source/common/thread_local/thread_local_impl.cc
index 14bcf251bd5c6..6884aae42d9a7 100644
--- a/source/common/thread_local/thread_local_impl.cc
+++ b/source/common/thread_local/thread_local_impl.cc
@@ -98,15 +98,15 @@ void InstanceImpl::runOnAllThreads(Event::PostCb cb, Event::PostCb all_threads_c
   // all_threads_complete_cb method. Parallelism of main thread execution is being traded off
   // for programming simplicity here.
   cb();
-  std::shared_ptr<std::atomic<uint64_t>> worker_count =
-      std::make_shared<std::atomic<uint64_t>>(registered_threads_.size());
+
+  std::shared_ptr<Event::PostCb> cb_guard(new Event::PostCb(cb),
+                                          [this, all_threads_complete_cb](Event::PostCb* cb) {
+                                            main_thread_dispatcher_->post(all_threads_complete_cb);
+                                            delete cb;
+                                          });
+
   for (Event::Dispatcher& dispatcher : registered_threads_) {
-    dispatcher.post([this, worker_count, cb, all_threads_complete_cb]() -> void {
-      cb();
-      if (--*worker_count == 0) {
-        main_thread_dispatcher_->post(all_threads_complete_cb);
-      }
-    });
+    dispatcher.post([cb_guard]() -> void { (*cb_guard)(); });
   }
 }