diff --git a/.github/scripts/recover-demo-runtime.sh b/.github/scripts/recover-demo-runtime.sh
index 13b266ec1d..132c5cf48f 100755
--- a/.github/scripts/recover-demo-runtime.sh
+++ b/.github/scripts/recover-demo-runtime.sh
@@ -32,6 +32,10 @@ HISTORY_BACKUP_ARCHIVE=""
 DEMO_HISTORY_PATHS=(
   ai_incidents.json
   ai_incidents.json.tmp
+  alerts/alert-history.json
+  alerts/alert-history.backup.json
+  alerts/alert-history.json.imported
+  alerts/alert-history.backup.json.imported
   alerts/events.db
   alerts/events.db-shm
   alerts/events.db-wal
@@ -348,7 +352,8 @@ sudo chmod 600 /etc/pulse/.env
 # operational memory. A failed 50-node bootstrap can leave thousands of
 # lifecycle events whose v6.4 replay queues one serialized incident save per
 # event before HTTP opens. Back up and clear only those fixed demo-history
-# files; compensation restores them if the bounded profile still fails.
+# files, including the retired JSON sources that repopulate a cleared event
+# log; compensation restores them if the bounded profile still fails.
 clear_demo_operational_history
 sudo systemctl restart "$SERVICE_NAME"
 
diff --git a/docs/release-control/v6/internal/subsystems/deployment-installability.md b/docs/release-control/v6/internal/subsystems/deployment-installability.md
index 0a244f738e..f55ac5d014 100644
--- a/docs/release-control/v6/internal/subsystems/deployment-installability.md
+++ b/docs/release-control/v6/internal/subsystems/deployment-installability.md
@@ -2054,7 +2054,9 @@ left more than 5,000 `IncidentStore.saveAsync` goroutines contending on one
 persistence mutex while durable mock alert lifecycle events were replayed.
 Stable-demo recovery therefore archives and clears only the fixed generated
 demo operational-memory files (`ai_incidents.json` and `alerts/events.db` plus
-SQLite sidecars) before restart. A failed bounded restart restores that archive;
+SQLite sidecars), together with the fixed retired `alert-history` JSON recovery
+sources that otherwise repopulate the cleared event log, before restart. A
+failed bounded restart restores that archive;
 success discards it. Customer configuration, active-alert recovery state,
 metrics history, the binary, unit, and Relay remain outside this reset.
 
diff --git a/scripts/release_control/release_promotion_policy_test.py b/scripts/release_control/release_promotion_policy_test.py
index d8e3c8e005..fe3a1cfd72 100644
--- a/scripts/release_control/release_promotion_policy_test.py
+++ b/scripts/release_control/release_promotion_policy_test.py
@@ -2128,6 +2128,8 @@ class ReleasePromotionPolicyTest(unittest.TestCase):
         self.assertIn("clear_demo_operational_history", recovery)
         self.assertIn("restore_demo_operational_history", recovery)
         self.assertIn("alerts/events.db", recovery)
+        self.assertIn("alerts/alert-history.json.imported", recovery)
+        self.assertIn("alerts/alert-history.backup.json.imported", recovery)
         self.assertIn("ai_incidents.json", recovery)
         self.assertIn("demo-operational-history.tar", recovery)
         self.assertNotIn('rm -f "/etc/pulse/', recovery)