diff --git a/js/package.json b/js/package.json
index c3516a19b..c696d4360 100644
--- a/js/package.json
+++ b/js/package.json
@@ -27,6 +27,7 @@
     "css-loader": "^3.0.0",
     "file-loader": "^4.0.0",
     "npm-run-all": "^4.1.5",
+    "promise-semaphore": "^0.2.8",
     "style-loader": "^0.23.1",
     "url-loader": "^1.0.0",
     "webpack": "^4.29.3",
diff --git a/js/src/manager.js b/js/src/manager.js
index 40aa79eaf..711fc3e22 100644
--- a/js/src/manager.js
+++ b/js/src/manager.js
@@ -16,6 +16,7 @@ import * as controls from '@jupyter-widgets/controls';
 import * as PhosphorWidget from '@phosphor/widgets';
 
 import { requireLoader } from './loader';
+import { batchRateMap } from './utils';
 
 if (typeof window !== "undefined" && typeof window.define !== "undefined") {
     window.define("@jupyter-widgets/base", base);
@@ -117,10 +118,17 @@ export class WidgetManager extends JupyterLabManager {
     async _build_models() {
         const comm_ids = await this._get_comm_info();
         const models = {};
-        const widgets_info = await Promise.all(Object.keys(comm_ids).map(async (comm_id) => {
+        /**
+         * For the classical notebook, iopub_msg_rate_limit=1000 (default)
+         * And for zmq, we are affected by the default ZMQ_SNDHWM setting of 1000
+         * See https://github.com/voila-dashboards/voila/issues/534 for a discussion
+         */
+        const maxMessagesInTransit = 100; // really save limit compared to ZMQ_SNDHWM
+        const maxMessagesPerSecond = 500; // lets be on the save side, in case the kernel sends more msg'es
+        const widgets_info = await batchRateMap(Object.keys(comm_ids), async (comm_id) => {
             const comm = await this._create_comm(this.comm_target_name, comm_id);
             return this._update_comm(comm);
-        }));
+        }, {room: maxMessagesInTransit, rate: maxMessagesPerSecond});
 
         await Promise.all(widgets_info.map(async (widget_info) => {
             const state = widget_info.msg.content.data.state;