diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c
index 251b9fe..46fe94d 100644
--- a/src/backend/postmaster/autovacuum.c
+++ b/src/backend/postmaster/autovacuum.c
@@ -1898,6 +1898,8 @@ do_autovacuum(void)
 	ScanKeyData key;
 	TupleDesc	pg_class_desc;
 	int			effective_multixact_freeze_max_age;
+	bool		did_vacuum = false;
+	bool		found_concurrent_worker = false;
 
 	/*
 	 * StartTransactionCommand and CommitTransactionCommand will automatically
@@ -2307,6 +2309,7 @@ do_autovacuum(void)
 			if (worker->wi_tableoid == relid)
 			{
 				skipit = true;
+				found_concurrent_worker = true;
 				break;
 			}
 		}
@@ -2433,6 +2436,8 @@ do_autovacuum(void)
 		}
 		PG_END_TRY();
 
+		did_vacuum = true;
+
 		/* the PGXACT flags are reset at the next end of transaction */
 
 		/* be tidy */
@@ -2470,8 +2475,20 @@ deleted:
 	/*
 	 * Update pg_database.datfrozenxid, and truncate pg_clog if possible. We
 	 * only need to do this once, not after each table.
+	 *
+	 * This action effectively causes launch of another auto-vacuum iteration,
+	 * hoping to choose another database. Now if this worker hasn't vaccuumed
+	 * anything because of another working already having taken one or more
+	 * tables, then there's no point going through another auto-vacuum
+	 * iteration: the other worker will do that job. In fact, if we do create
+	 * another iteration, we might keep spawning useless autovacuum iterations
+	 * on the same database, until the other worker has finished with its
+	 * tables and updated datfrozenxid.
 	 */
-	vac_update_datfrozenxid();
+	if (!did_vacuum && found_concurrent_worker)
+		; /* Don't do anything */
+	else
+		vac_update_datfrozenxid();
 
 	/* Finally close out the last transaction. */
 	CommitTransactionCommand();