v2: fix agent-runner lifecycle and session DB reliability

- Use DELETE journal mode for session DBs instead of WAL. WAL doesn't
  sync reliably across Docker volume mounts (VirtioFS), causing dropped
  writes and duplicate deliveries.
- Add 20s idle detection to end the query stream. The concurrent poll
  tracks SDK activity via a new 'activity' provider event. When no SDK
  events arrive for 20s and no messages are pending, the stream ends
  and the poll loop continues.
- Add touchProcessing heartbeat so the host can distinguish active
  agents from idle ones by checking status_changed recency.
- Catch query errors in the poll loop and write error responses to
  messages_out instead of crashing the process.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

container/agent-runner/src/db/connection.ts

@@ -7,7 +7,7 @@ let _db: Database.Database | null = null;
 export function getSessionDb(): Database.Database {
   if (!_db) {
     _db = new Database(process.env.SESSION_DB_PATH || SESSION_DB_PATH);
-    _db.pragma('journal_mode = WAL');
+    _db.pragma('journal_mode = DELETE');
     _db.pragma('foreign_keys = ON');
   }
   return _db;

container/agent-runner/src/db/messages-in.ts

@@ -47,6 +47,14 @@ export function markCompleted(ids: string[]): void {
   })();
 }
 
+/** Update status_changed on processing messages (heartbeat for host idle detection). */
+export function touchProcessing(ids: string[]): void {
+  if (ids.length === 0) return;
+  const db = getSessionDb();
+  const stmt = db.prepare("UPDATE messages_in SET status_changed = datetime('now') WHERE id = ? AND status = 'processing'");
+  for (const id of ids) stmt.run(id);
+}
+
 /** Mark a single message as failed. */
 export function markFailed(id: string): void {
   getSessionDb().prepare("UPDATE messages_in SET status = 'failed', status_changed = datetime('now') WHERE id = ?").run(id);