fix: enforce one Pi runtime per user

backend/src/routes/sessions.ts

@@ -97,14 +97,6 @@ export function sessionRoutes(
     return true;
   };
 
-  const releaseFinalizedRuntimes = async (): Promise<void> => {
-    await Promise.all([...boundRuntimes.entries()].map(([id, rt]) =>
-      withSessionLifecycle(id, () => releaseIfFinalized(id, rt)).catch((error: unknown) => {
-        console.error(`[session:${id}] stale runtime cleanup failed:`, error);
-      }),
-    ));
-  };
-
   const bindRuntime = (
     id: string, rt: ReturnType<PiProcessManager["createFor"]>, runner: any, workspace?: string,
   ) => {
@@ -208,7 +200,10 @@ export function sessionRoutes(
     let s: Settings;
     try { s = await d.getSettings(principal); } catch { return storageFailure(reply); }
     const runner = runnerFor(principal);
-    await releaseFinalizedRuntimes();
+    // A persisted session is resumable without keeping Pi alive. New work replaces every
+    // runtime owned by this principal, while runtimes belonging to other users remain intact.
+    // Optional chaining preserves the deliberately narrow manager stubs used by route tests.
+    for (const id of d.mgr.teardownForPrincipal?.(principal) ?? []) boundRuntimes.delete(id);
     const ensure = await d.readiness.ensure(s.workspace ?? "", principal);
     if (!ensure.ok) return reply.code(503).send({ error: READINESS_FAILURE_MESSAGE });
     // Local-only: verify the DWH is reachable BEFORE creating the session, so a dropped
@@ -388,6 +383,10 @@ export function sessionRoutes(
         }
       }
 
+      for (const stoppedId of d.mgr.teardownForPrincipal?.(principal) ?? []) {
+        boundRuntimes.delete(stoppedId);
+      }
+
       let rt: ReturnType<PiProcessManager["createFor"]> | undefined;
       try {
         if (current) {