diff --git a/code/object/objcollide.cpp b/code/object/objcollide.cpp
index 5e11d15a3ed..4ece1192746 100644
--- a/code/object/objcollide.cpp
+++ b/code/object/objcollide.cpp
@@ -762,6 +762,7 @@ void spin_up_mp_collision() {
 void spin_down_mp_collision() {
 	threading::spin_down_threaded_task();
 	collision_processing_done.store(true);
+	threading::spin_down_wait_complete();
 }
 
 void queue_mp_collision(uint ctype, const obj_pair& colliding) {
diff --git a/code/utils/threading.cpp b/code/utils/threading.cpp
index 5c71228235e..27787263021 100644
--- a/code/utils/threading.cpp
+++ b/code/utils/threading.cpp
@@ -20,6 +20,7 @@ namespace threading {
 	static std::condition_variable wait_for_task;
 	static std::mutex wait_for_task_mutex;
 	static bool wait_for_task_condition;
+	static std::atomic_uint32_t wait_for_spindown_tasks;
 	static std::atomic<WorkerThreadTask> worker_task;
 
 	static SCP_vector<std::thread> worker_threads;
@@ -28,12 +29,15 @@ namespace threading {
 	static void mp_worker_thread_main(size_t threadIdx) {
 		while(true) {
 			{
+				//We're waiting for a new task, so spindown was successful
+				wait_for_spindown_tasks.fetch_add(1, std::memory_order_release);
 				std::unique_lock<std::mutex> lk(wait_for_task_mutex);
 				wait_for_task.wait(lk, []() { return wait_for_task_condition; });
 			}
 
 			switch (worker_task.load(std::memory_order_acquire)) {
 				case WorkerThreadTask::EXIT:
+					//We're done and will quit, so ensure we report this.
 					return;
 				case WorkerThreadTask::COLLISION:
 					collide_mp_worker_thread(threadIdx);
@@ -144,6 +148,7 @@ namespace threading {
 	//External Functions
 
 	void spin_up_threaded_task(WorkerThreadTask task) {
+		wait_for_spindown_tasks.store(0, std::memory_order_release);
 		worker_task.store(task);
 		{
 			std::scoped_lock lock {wait_for_task_mutex};
@@ -157,6 +162,11 @@ namespace threading {
 		wait_for_task_condition = false;
 	}
 
+	void spin_down_wait_complete() {
+		//Technically, spindowns should only occur when the actual code is confirmed to be complete. So busy-waiting here is not an issue.
+		while (wait_for_spindown_tasks.load(std::memory_order_acquire) < num_threads);
+	}
+
 	void init_task_pool() {
 		if (Cmdline_multithreading == 0) {
 			//At least given the current collision-detection threading, 8 cores (if available) seems like a sweetspot, with more cores adding too much overhead.
@@ -181,6 +191,8 @@ namespace threading {
 	void shut_down_task_pool() {
 		spin_up_threaded_task(WorkerThreadTask::EXIT);
 
+		//Technically we could await spin_down_wait_complete here, but since we're returning and joining the threads here, there is no need
+
 		for(auto& thread : worker_threads) {
 			thread.join();
 		}
diff --git a/code/utils/threading.h b/code/utils/threading.h
index ab33e655bbd..5da628ec114 100644
--- a/code/utils/threading.h
+++ b/code/utils/threading.h
@@ -11,6 +11,9 @@ namespace threading {
 	//This _must_ be called on the main thread BEFORE a task completes on a thread of the task pool.
 	void spin_down_threaded_task();
 
+	//This should be called AFTER the command to finish a given task is given. This will block until all threads have returned into a state where they are able to listen to new commands.
+	void spin_down_wait_complete();
+
 	void init_task_pool();
 	void shut_down_task_pool();