diff --git a/package.json b/package.json
index a174deadcd..951354bd75 100644
--- a/package.json
+++ b/package.json
@@ -148,6 +148,7 @@
     "multihashes": "~0.4.14",
     "multihashing-async": "~0.6.0",
     "node-fetch": "^2.3.0",
+    "p-queue": "^6.1.0",
     "peer-book": "~0.9.0",
     "peer-id": "~0.12.3",
     "peer-info": "~0.15.0",
diff --git a/src/core/runtime/dns-browser.js b/src/core/runtime/dns-browser.js
index 74d0a14cdb..7b2725e491 100644
--- a/src/core/runtime/dns-browser.js
+++ b/src/core/runtime/dns-browser.js
@@ -1,33 +1,57 @@
-/* global self */
+/* eslint-env browser */
 'use strict'
 
+const TLRU = require('../../utils/tlru')
+const { default: PQueue } = require('p-queue')
+
+// Avoid sending multiple queries for the same hostname by caching results
+const cache = new TLRU(1000)
+// TODO: /api/v0/dns does not return TTL yet: https://github.com/ipfs/go-ipfs/issues/5884
+// However we know browsers themselves cache DNS records for at least 1 minute,
+// which acts a provisional default ttl: https://stackoverflow.com/a/36917902/11518426
+const ttl = 60 * 1000
+
+// browsers limit concurrent connections per host,
+// we don't want preload calls to exhaust the limit (~6)
+const _httpQueue = new PQueue({ concurrency: 4 })
+
+function unpackResponse (domain, response, callback) {
+  if (response.Path) {
+    return callback(null, response.Path)
+  } else {
+    const err = new Error(response.Message)
+    return callback(err)
+  }
+}
+
 module.exports = (domain, opts, callback) => {
   if (typeof opts === 'function') {
     callback = opts
     opts = {}
   }
-
   opts = opts || {}
-
   domain = encodeURIComponent(domain)
-  let url = `https://ipfs.io/api/v0/dns?arg=${domain}`
 
+  if (cache.has(domain)) {
+    const response = cache.get(domain)
+    return unpackResponse(domain, response, callback)
+  }
+
+  let url = `https://ipfs.io/api/v0/dns?arg=${domain}`
   Object.keys(opts).forEach(prop => {
     url += `&${encodeURIComponent(prop)}=${encodeURIComponent(opts[prop])}`
   })
 
-  self.fetch(url, { mode: 'cors' })
+  _httpQueue.add(() => fetch(url, { mode: 'cors' })
     .then((response) => {
       return response.json()
     })
     .then((response) => {
-      if (response.Path) {
-        return callback(null, response.Path)
-      } else {
-        return callback(new Error(response.Message))
-      }
+      cache.set(domain, response, ttl)
+      return unpackResponse(domain, response, callback)
     })
     .catch((error) => {
       callback(error)
     })
+  )
 }
diff --git a/src/core/runtime/preload-browser.js b/src/core/runtime/preload-browser.js
index 81407f483f..d24c7f3eff 100644
--- a/src/core/runtime/preload-browser.js
+++ b/src/core/runtime/preload-browser.js
@@ -1,18 +1,23 @@
 /* eslint-env browser */
 'use strict'
 
+const { default: PQueue } = require('p-queue')
 const debug = require('debug')
 
 const log = debug('ipfs:preload')
 log.error = debug('ipfs:preload:error')
 
+// browsers limit concurrent connections per host,
+// we don't want preload calls to exhaust the limit (~6)
+const _httpQueue = new PQueue({ concurrency: 4 })
+
 module.exports = function preload (url, callback) {
   log(url)
 
   const controller = new AbortController()
   const signal = controller.signal
 
-  fetch(url, { signal })
+  _httpQueue.add(() => fetch(url, { signal })
     .then(res => {
       if (!res.ok) {
         log.error('failed to preload', url, res.status, res.statusText)
@@ -22,6 +27,7 @@ module.exports = function preload (url, callback) {
     })
     .then(() => callback())
     .catch(callback)
+  ).catch(callback)
 
   return {
     cancel: () => controller.abort()