diff --git a/docs/en/operations.md b/docs/en/operations.md
index 47b6bfa2..f5e62823 100644
--- a/docs/en/operations.md
+++ b/docs/en/operations.md
@@ -49,6 +49,8 @@ reme reindex scope=embedding
 
 `reindex` rebuilds BM25 and/or embedding indexes from the current `file_chunks`. It does not scan the workspace, rechunk files, or rebuild the wikilink graph. Diagnose the watcher first when ingestion is the problem.
 
+When a ReMe upgrade or chunker configuration change alters chunking behavior, existing chunks do not refresh automatically. To apply the new behavior, follow [Recover derived state](#recover-derived-state).
+
 Rebuild a daily index page separately:
 
 ```bash
@@ -90,9 +92,10 @@ Do not remove anything until a backup exists. Then:
 1. preserve `session/`, `resource/`, `daily/`, and `digest/`;
 2. record the effective configuration and component backends;
 3. verify that the failure is limited to `metadata/`;
-4. move suspect derived state to an isolated backup location;
-5. restart with the same configuration and let watchers rebuild;
-6. validate search, graph traversal, and daily indexes.
+4. stop the service, so shutdown persistence cannot write the old in-memory state back over the cleanup;
+5. move only the affected store's derived state to an isolated backup location, preserving unrelated catalogs;
+6. restart with the same configuration and let watchers rebuild the files in their configured watch scope;
+7. validate search, graph traversal, and daily indexes.
 
 The internal layout of metadata files is not a public automation contract.
 
diff --git a/docs/zh/operations.md b/docs/zh/operations.md
index e66b7055..27571bf9 100644
--- a/docs/zh/operations.md
+++ b/docs/zh/operations.md
@@ -52,6 +52,8 @@ reme reindex scope=embedding
 
 如果问题发生在文件摄取阶段，应先确认后台 watcher 正常运行；不能把 `reindex` 当成通用“重新扫描”命令。
 
+升级 ReMe 或修改 chunker 配置导致分块规则变化后，已有分块不会自动刷新。如需应用新规则，请按[从派生状态故障恢复](#从派生状态故障恢复)流程重建。
+
 Daily 索引页可单独重建：
 
 ```bash
@@ -93,9 +95,10 @@ reme start workspace_dir=/new/location/reme-memory
 1. 保留 `session/`、`resource/`、`daily/` 和 `digest/`；
 2. 记录当前配置和组件 backend；
 3. 确认故障只发生在 `metadata/`；
-4. 将有问题的派生状态移到隔离备份位置；
-5. 用同一配置启动 ReMe，让 watcher 从源文件重建；
-6. 验证搜索、图谱和 Daily 索引。
+4. 先停止服务，避免关闭时的持久化把旧的内存状态写回、覆盖刚做的清理；
+5. 只把受影响存储的派生状态移到隔离备份位置，保留无关 catalog 状态；
+6. 用同一配置启动 ReMe，让 watcher 重建其配置覆盖的文件；
+7. 验证搜索、图谱和 Daily 索引。
 
 具体 metadata 文件属于实现细节，不应在自动化脚本里依赖其内部格式。