deploy: start-first updates to reduce downtime

stack.yml

@@ -38,8 +38,11 @@ services:
       replicas: 1
       update_config:
         parallelism: 1
-        order: stop-first
+        # start-first evita downtime: sobe o novo task antes de parar o anterior
+        order: start-first
         failure_action: rollback
+      rollback_config:
+        order: start-first
       resources:
         limits:
           memory: "2G"