docs(js): Add section on tracing at high concurrency in serverless en…

…vs (#611)

docs/observability/how_to_guides/tracing/serverless_environments.mdx

@@ -35,3 +35,39 @@ const res = await tracedFn();
 
 await langsmithClient.awaitPendingTraceBatches();
 ```
+
+## Rate limits at high concurrency
+
+By default, the LangSmith client will batch operations as your traced run executions, sending a new batch every few milliseconds.
+
+This works well in most situations, but if your traced function is long-running and you have very high concurrency,
+you may also hit rate limits related to overall request count.
+
+If you are seeing rate limit errors related to this, you can try setting `manualFlushMode: true` in your client like this:
+
+```ts
+import { Client } from "langsmith";
+
+const langsmithClient = new Client({
+  manualFlushMode: true,
+});
+
+const myTracedFunc = traceable(
+  async () => {
+    // Your logic here...
+  },
+  { client: langsmithClient }
+);
+```
+
+And then manually calling `client.flush()` like this before your serverless function closes:
+
+```ts
+try {
+  await myTracedFunc();
+} finally {
+  await langsmithClient.flush();
+}
+```
+
+Note that this will prevent runs from appearing in the LangSmith UI until you call `.flush()`.