Fix a test that is falsely classified as flaky

This test is written in such a way that it fails when run consecutively.
The test assumes that the second worker nodes does not have metadata,
but after the whole test file is executed, the worker has the metadata.
Therefore all the runs after the first one fails.

src/test/regress/expected/multi_unsupported_worker_operations.out

@@ -157,6 +157,13 @@ SELECT master_remove_node('localhost', 5432);
 
 (1 row)
 
+SELECT stop_metadata_sync_to_node('localhost', :worker_2_port);
+NOTICE:  dropping metadata on the node (localhost,57638)
+ stop_metadata_sync_to_node
+---------------------------------------------------------------------
+
+(1 row)
+
 \c - - - :worker_1_port
 UPDATE pg_dist_partition SET colocationid = 0 WHERE logicalrelid='mx_table_2'::regclass;
 SELECT update_distributed_table_colocation('mx_table', colocate_with => 'mx_table_2');
@@ -181,7 +188,7 @@ HINT:  Connect to the coordinator and run it again.
 SELECT hasmetadata FROM pg_dist_node WHERE nodeport=:worker_2_port;
  hasmetadata
 ---------------------------------------------------------------------
- t
+ f
 (1 row)
 
 -- stop_metadata_sync_to_node

src/test/regress/sql/multi_unsupported_worker_operations.sql

@@ -96,6 +96,7 @@ SELECT count(1) FROM pg_dist_node WHERE nodename='localhost' AND nodeport=5432;
 
 \c - - - :master_port
 SELECT master_remove_node('localhost', 5432);
+SELECT stop_metadata_sync_to_node('localhost', :worker_2_port);
 
 \c - - - :worker_1_port