diff --git a/.github/workflows/build-and-push-bsky-aws.yaml b/.github/workflows/build-and-push-bsky-aws.yaml
index beaf10eb655..34bba3070cd 100644
--- a/.github/workflows/build-and-push-bsky-aws.yaml
+++ b/.github/workflows/build-and-push-bsky-aws.yaml
@@ -3,7 +3,7 @@ on:
   push:
     branches:
       - main
-      - timeline-limit-1-opt
+      - bsky-node-clustering
 env:
   REGISTRY: ${{ secrets.AWS_ECR_REGISTRY_USEAST2_PACKAGES_REGISTRY }}
   USERNAME: ${{ secrets.AWS_ECR_REGISTRY_USEAST2_PACKAGES_USERNAME }}
diff --git a/services/bsky/api.js b/services/bsky/api.js
index c4882335761..42737d72b56 100644
--- a/services/bsky/api.js
+++ b/services/bsky/api.js
@@ -13,6 +13,7 @@ require('dd-trace') // Only works with commonjs
 // Tracer code above must come before anything else
 const path = require('path')
 const assert = require('assert')
+const cluster = require('cluster')
 const {
   BunnyInvalidator,
   CloudfrontInvalidator,
@@ -140,12 +141,14 @@ const main = async () => {
 
   await bsky.start()
   // Graceful shutdown (see also https://aws.amazon.com/blogs/containers/graceful-shutdowns-with-ecs/)
-  process.on('SIGTERM', async () => {
+  const shutdown = async () => {
     // Gracefully shutdown periodic-moderation-event-reversal before destroying bsky instance
     periodicModerationEventReversal.destroy()
     await periodicModerationEventReversalRunning
     await bsky.destroy()
-  })
+  }
+  process.on('SIGTERM', shutdown)
+  process.on('disconnect', shutdown) // when clustering
 }
 
 const getEnv = () => ({
@@ -223,4 +226,31 @@ const maintainXrpcResource = (span, req) => {
   }
 }
 
-main()
+const workerCount = maybeParseInt(process.env.CLUSTER_WORKER_COUNT)
+
+if (workerCount) {
+  if (cluster.isPrimary) {
+    console.log(`primary ${process.pid} is running`)
+    const workers = new Set()
+    for (let i = 0; i < workerCount; ++i) {
+      workers.add(cluster.fork())
+    }
+    let teardown = false
+    cluster.on('exit', (worker) => {
+      workers.delete(worker)
+      if (!teardown) {
+        workers.add(cluster.fork()) // restart on crash
+      }
+    })
+    process.on('SIGTERM', () => {
+      teardown = true
+      console.log('disconnecting workers')
+      workers.forEach((w) => w.disconnect())
+    })
+  } else {
+    console.log(`worker ${process.pid} is running`)
+    main()
+  }
+} else {
+  main() // non-clustering
+}