diff --git a/roles/openqa/worker/files/kill-stuck-qemu.sh b/roles/openqa/worker/files/kill-stuck-qemu.sh
new file mode 100755
index 0000000000..05e0a0c61b
--- /dev/null
+++ b/roles/openqa/worker/files/kill-stuck-qemu.sh
@@ -0,0 +1,7 @@
+#!/bin/bash
+
+# this is a hideous hack to find and kill qemu processes stuck as a
+# result of https://github.com/os-autoinst/os-autoinst/issues/2549
+# which cause workers to be stuck in broken state. affected workers
+# should recover some minutes after this script runs
+for i in {1..35}; do journalctl -u openqa-worker-plain@$i.service -n 5 | grep "is still running" | grep -o "PID: [0-9]\+" | cut -d" " -f2 | sort -u | xargs kill 2> /dev/null; done
diff --git a/roles/openqa/worker/tasks/main.yml b/roles/openqa/worker/tasks/main.yml
index ad5f4088d1..713d0b1af3 100644
--- a/roles/openqa/worker/tasks/main.yml
+++ b/roles/openqa/worker/tasks/main.yml
@@ -167,6 +167,9 @@
   service: name=rngd enabled=yes state=started
   when: "openqa_rngd is defined and openqa_rngd"
 
+- name: Install cron job to kill stuck qemu processes
+  copy: src=kill-stuck-qemu.sh dest=/etc/cron.daily/kill-stuck-qemu owner=root group=root mode=0755
+
 - include_tasks: nfs-client.yml
   when: openqa_nfs_worker|bool