diff --git a/src/backend/distributed/executor/multi_real_time_executor.c b/src/backend/distributed/executor/multi_real_time_executor.c
index ca1a40c62..c13c41a78 100644
--- a/src/backend/distributed/executor/multi_real_time_executor.c
+++ b/src/backend/distributed/executor/multi_real_time_executor.c
@@ -387,7 +387,7 @@ ManageTaskExecution(Task *task, TaskExecution *taskExecution,
 
 		case EXEC_TASK_FAILED:
 		{
-			bool raiseError = true;
+			bool raiseError = false;
 
 			/*
 			 * On task failure, we close the connection. We also reset our execution
@@ -397,12 +397,15 @@ ManageTaskExecution(Task *task, TaskExecution *taskExecution,
 			 */
 			int32 connectionId = connectionIdArray[currentIndex];
 			MultiConnection *connection = MultiClientGetConnection(connectionId);
+			bool isCritical = connection->remoteTransaction.transactionCritical;
 
 			/*
-			 * If this connection was previously marked as critical (e.g. it was used
-			 * to perform a DDL command), then throw an error. Otherwise, mark it
-			 * as failed and continue executing the query.
+			 * Mark the connection as failed in case it was already used to perform
+			 * writes. We do not error out here, because the abort handler currently
+			 * cannot handle connections with COPY (SELECT ...) TO STDOUT commands
+			 * in progress.
 			 */
+			raiseError = false;
 			MarkRemoteTransactionFailed(connection, raiseError);
 
 			MultiClientDisconnect(connectionId);
@@ -411,8 +414,16 @@ ManageTaskExecution(Task *task, TaskExecution *taskExecution,
 
 			taskStatusArray[currentIndex] = EXEC_TASK_CONNECT_START;
 
-			/* try next worker node */
-			AdjustStateForFailure(taskExecution);
+			if (isCritical)
+			{
+				/* cannot recover when error occurs in a critical transaction */
+				taskExecution->criticalErrorOccurred = true;
+			}
+			else
+			{
+				/* try next worker node */
+				AdjustStateForFailure(taskExecution);
+			}
 
 			/*
 			 * Add a delay, to avoid potentially excerbating problems by
diff --git a/src/backend/distributed/executor/multi_server_executor.c b/src/backend/distributed/executor/multi_server_executor.c
index b1196eff2..46d408fe8 100644
--- a/src/backend/distributed/executor/multi_server_executor.c
+++ b/src/backend/distributed/executor/multi_server_executor.c
@@ -249,6 +249,11 @@ CleanupTaskExecution(TaskExecution *taskExecution)
 bool
 TaskExecutionFailed(TaskExecution *taskExecution)
 {
+	if (taskExecution->criticalErrorOccurred)
+	{
+		return true;
+	}
+
 	if (taskExecution->failureCount >= MAX_TASK_EXECUTION_FAILURES)
 	{
 		return true;
diff --git a/src/include/distributed/multi_server_executor.h b/src/include/distributed/multi_server_executor.h
index 6f2ef0c28..66e5dbec6 100644
--- a/src/include/distributed/multi_server_executor.h
+++ b/src/include/distributed/multi_server_executor.h
@@ -145,6 +145,7 @@ struct TaskExecution
 	uint32 querySourceNodeIndex; /* only applies to map fetch tasks */
 	int32 dataFetchTaskIndex;
 	uint32 failureCount;
+	bool criticalErrorOccurred;
 };
 
 
diff --git a/src/test/regress/expected/subquery_and_cte.out b/src/test/regress/expected/subquery_and_cte.out
index 1a785f9ff..e2c63a7a2 100644
--- a/src/test/regress/expected/subquery_and_cte.out
+++ b/src/test/regress/expected/subquery_and_cte.out
@@ -435,6 +435,42 @@ FROM
      ) as bar  
 WHERE foo.user_id = bar.user_id;
 ERROR:  recursive CTEs are not supported in distributed queries
+-- We error-out when there's an error in execution of the query. By repeating it
+-- multiple times, we increase the chance of this test failing before PR #1903.
+SET client_min_messages TO ERROR;
+DO $$
+DECLARE
+	errors_received INTEGER;
+BEGIN
+errors_received := 0;
+FOR i IN 1..3 LOOP
+	BEGIN
+		WITH cte as (
+			SELECT
+				user_id, value_2
+			from
+				events_table
+		)
+		SELECT * FROM users_table where value_2 < (
+			SELECT
+				min(cte.value_2)
+			FROM
+				cte
+			WHERE
+				users_table.user_id=cte.user_id
+			GROUP BY
+				user_id, cte.value_2);
+	EXCEPTION WHEN OTHERS THEN
+		IF SQLERRM LIKE 'failed to execute task%' THEN
+			errors_received := errors_received + 1;
+		END IF;
+	END;
+END LOOP;
+RAISE '(%/3) failed to execute one of the tasks', errors_received;
+END;
+$$;
+ERROR:  (3/3) failed to execute one of the tasks
+CONTEXT:  PL/pgSQL function inline_code_block line 29 at RAISE
 SET client_min_messages TO DEFAULT;
 DROP SCHEMA subquery_and_ctes CASCADE;
 NOTICE:  drop cascades to table users_table_local
diff --git a/src/test/regress/sql/subquery_and_cte.sql b/src/test/regress/sql/subquery_and_cte.sql
index 7636a921b..3ec23cb39 100644
--- a/src/test/regress/sql/subquery_and_cte.sql
+++ b/src/test/regress/sql/subquery_and_cte.sql
@@ -319,6 +319,40 @@ FROM
      ) as bar  
 WHERE foo.user_id = bar.user_id;
 
+-- We error-out when there's an error in execution of the query. By repeating it
+-- multiple times, we increase the chance of this test failing before PR #1903.
+SET client_min_messages TO ERROR;
+DO $$
+DECLARE
+	errors_received INTEGER;
+BEGIN
+errors_received := 0;
+FOR i IN 1..3 LOOP
+	BEGIN
+		WITH cte as (
+			SELECT
+				user_id, value_2
+			from
+				events_table
+		)
+		SELECT * FROM users_table where value_2 < (
+			SELECT
+				min(cte.value_2)
+			FROM
+				cte
+			WHERE
+				users_table.user_id=cte.user_id
+			GROUP BY
+				user_id, cte.value_2);
+	EXCEPTION WHEN OTHERS THEN
+		IF SQLERRM LIKE 'failed to execute task%' THEN
+			errors_received := errors_received + 1;
+		END IF;
+	END;
+END LOOP;
+RAISE '(%/3) failed to execute one of the tasks', errors_received;
+END;
+$$;
 
 SET client_min_messages TO DEFAULT;