diff --git a/journal_upload/handlers/main.yml b/journal_upload/handlers/main.yml
index bd06f75..ea6d0b5 100644
--- a/journal_upload/handlers/main.yml
+++ b/journal_upload/handlers/main.yml
@@ -1,4 +1,8 @@
 ---
+- name: Reload systemd manager
+  ansible.builtin.systemd:
+    daemon_reload: true
+
 - name: Restart systemd-journal-upload
   ansible.builtin.systemd:
     name: systemd-journal-upload
diff --git a/journal_upload/tasks/main.yml b/journal_upload/tasks/main.yml
index 89604c5..f8efbaf 100644
--- a/journal_upload/tasks/main.yml
+++ b/journal_upload/tasks/main.yml
@@ -49,6 +49,31 @@
   notify: Restart systemd-journal-upload
   no_log: true
 
+# The default start limit (5 starts within 10 seconds) gives up during an ingest outage. The
+# uploader resumes from its cursor, so it keeps retrying.
+- name: Create the uploader unit drop-in directory
+  ansible.builtin.file:
+    path: /etc/systemd/system/systemd-journal-upload.service.d
+    state: directory
+    mode: "0755"
+
+- name: Keep the uploader retrying through ingest outages
+  ansible.builtin.copy:
+    dest: /etc/systemd/system/systemd-journal-upload.service.d/restart.conf
+    mode: "0644"
+    content: |
+      [Unit]
+      StartLimitIntervalSec=0
+
+      [Service]
+      Restart=always
+      RestartSec=30s
+  notify: Reload systemd manager
+
+# The reload has to precede the start below, which otherwise hits the default start limit.
+- name: Load the uploader drop-in
+  ansible.builtin.meta: flush_handlers
+
 - name: Enable the journal uploader
   ansible.builtin.systemd:
     name: systemd-journal-upload