diff --git a/create_pod_espocrm.sh b/create_pod_espocrm.sh
index 032b6c7..20749e7 100755
--- a/create_pod_espocrm.sh
+++ b/create_pod_espocrm.sh
@@ -103,6 +103,17 @@ for i in $(seq 1 90); do
 done
 echo "EspoCRM responded with HTTP $code (rc=$?)"
 
+# Rebuild EspoCRM cache so customizations on the persisted bind-mount take effect
+# and the client cacheTimestamp is bumped (forces browsers to reload custom views,
+# e.g. custom field views like the signature image button). The image entrypoint
+# does NOT rebuild on a normal same-version restart, so without this a recreate
+# keeps serving a stale client cache. Done here in the live phase where the DB is
+# already confirmed up (the systemd-managed phase has a start-up race on the DB).
+# The bumped cacheTimestamp and rebuilt cache persist on the bind-mount into the
+# systemd-managed instance.
+podman exec -u www-data "$WEB_CTR" php command.php rebuild
+echo "Rebuilt EspoCRM cache (rc=$?)"
+
 # EspoCRM scheduler/daemon container (start last; reads data/config.php)
 podman rm -f "$DAEMON_CTR"
 podman run -d --name "$DAEMON_CTR" --pod "$POD_NAME" \