fix: attribute fallback create failures and stop verifier before servers

Review follow-ups: the workload-create error log now carries the run id
(fallback creates run outside the dequeue wide event, so the log was the
only attribution), and the verifier stops before the workload server and
session so its timer can't cold-create a workload mid-shutdown.

Author: myftija

apps/supervisor/src/index.ts

@@ -606,7 +606,10 @@ class ManagedSupervisor {
         { backend: this.workloadManagerBackend, outcome: "error" },
         (performance.now() - createStart) / 1000
       );
-      this.logger.error("Failed to create workload", { error });
+      this.logger.error("Failed to create workload", {
+        runId: message.run.friendlyId,
+        error,
+      });
     }
   }
 
@@ -691,9 +694,11 @@ class ManagedSupervisor {
 
   async stop() {
     this.logger.log("Shutting down");
+    // Stop the verifier first: its timer can otherwise fire mid-shutdown and
+    // cold-create a workload on a node that is going down.
+    this.warmStartVerifier?.stop();
     await this.workloadServer.stop();
     await this.workerSession.stop();
-    this.warmStartVerifier?.stop();
 
     // Optional services
     this.backpressureMonitor?.stop();