Fix transient health checks and avoid unnecessary recovery

src/lib/server/provisioning/requests.ts

@@ -139,12 +139,21 @@ export function requestLifecycle(server: string, action: RuntimeAction) {
     return db().transaction(() => {
         const blocker = action === 'rebuild' ? activeBetaRemovalBlocker(server) : null;
         if (blocker) return { ok: false as const, status: 409, error: blocker };
-        const existing = db().prepare(`SELECT id, action FROM server_deployments WHERE server = ?
+        const existing = db().prepare(`SELECT id, action, state, submitted_at, triggered_by FROM server_deployments WHERE server = ?
             AND server_deleted_at IS NULL
-            AND state IN ('queued','waiting','running','awaiting-health') ORDER BY rowid DESC LIMIT 1`).get(server) as { id: string; action: string } | undefined;
+            AND state IN ('queued','waiting','running','awaiting-health') ORDER BY rowid DESC LIMIT 1`).get(server) as
+            { id: string; action: string; state: string; submitted_at: string | null; triggered_by: string } | undefined;
         if (existing) {
             if (existing.action !== action) return { ok: false as const, status: 409, error: 'Another provisioning operation is already pending for this server.' };
             db().prepare('UPDATE servers SET desired_state = ? WHERE name = ?').run(action === 'stop' ? 'stopped' : 'running', server);
+            if (existing.state === 'waiting' && !existing.submitted_at && existing.triggered_by.startsWith('watchdog:')
+                && db().prepare(`SELECT 1 FROM provisioning_sources s JOIN watchdog_incidents i ON i.id = s.source_id
+                    WHERE s.operation_id = ? AND s.kind = 'watchdog' AND i.kind = 'health'`).get(existing.id)) {
+                // An operator explicitly requesting this queued action keeps
+                // that intent even if the original health condition clears.
+                db().prepare("UPDATE server_deployments SET triggered_by = 'ui' WHERE id = ?").run(existing.id);
+                recordDeploymentEvent(existing.id, 'Operator explicitly requested the queued health recovery action');
+            }
             return { ok: true as const, deploymentId: existing.id };
         }
         const row = getServer(server);