fix_parallel_worker_startup_failures_v2.patch

diff --git a/src/backend/access/transam/parallel.c b/src/backend/access/transam/parallel.c
index d3431a7..5c7b794 100644
--- a/src/backend/access/transam/parallel.c
+++ b/src/backend/access/transam/parallel.c
@@ -538,8 +538,16 @@ LaunchParallelWorkers(ParallelContext *pcxt)
 void
 WaitForParallelWorkersToFinish(ParallelContext *pcxt)
 {
+	bool	   *alive_workers;
+	bool	   *dead_workers;
+	int			nworkers = pcxt->nworkers_launched;
+
+	alive_workers = (bool *) palloc0(nworkers * sizeof(bool));
+	dead_workers = (bool *) palloc0(nworkers * sizeof(bool));
+
 	for (;;)
 	{
+		BgwHandleStatus status;
 		bool		anyone_alive = false;
 		int			i;
 
@@ -550,9 +558,31 @@ WaitForParallelWorkersToFinish(ParallelContext *pcxt)
 		 */
 		CHECK_FOR_INTERRUPTS();
 
-		for (i = 0; i < pcxt->nworkers_launched; ++i)
+		for (i = 0; i < pcxt->nworkers; ++i)
 		{
-			if (pcxt->worker[i].error_mqh != NULL)
+			pid_t		pid;
+
+			if (pcxt->worker == NULL || pcxt->worker[i].bgwhandle == NULL)
+				continue;
+
+			/*
+			 * Check for unexpected worker death.  This will ensure that if
+			 * the postmaster failed to start the worker or the worker errored
+			 * out before attaching to error queue, then we don't wait for it
+			 * indefinitely.
+			 */
+			if (!alive_workers[i])
+			{
+				status = GetBackgroundWorkerPid(pcxt->worker[i].bgwhandle,
+												&pid);
+				if (status == BGWH_STOPPED)
+					dead_workers[i] = true;
+				if (status != BGWH_NOT_YET_STARTED)
+					alive_workers[i] = true;
+			}
+
+			if (pcxt->worker[i].error_mqh != NULL &&
+				!dead_workers[i])
 			{
 				anyone_alive = true;
 				break;
@@ -567,6 +597,10 @@ WaitForParallelWorkersToFinish(ParallelContext *pcxt)
 		ResetLatch(MyLatch);
 	}
 
+	/* be tidy */
+	pfree(alive_workers);
+	pfree(dead_workers);
+
 	if (pcxt->toc != NULL)
 	{
 		FixedParallelState *fps;