You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Bug]: Sidebar agent rows become permanently unclickable after the remote relay daemon restarts #9030
After the relay daemon on an SSH host restarts (crash or manual kill), agent rows for that project stay visible in the sidebar but clicking them does nothing — no tab opens, no toast, no console error. They stay like that until the 30-minute status freshness window expires.
Traced it in current main:
The rows are driven by agentStatusByPaneKey, which is cleared on explicit tab close or worktree sleep/remove but never on relay/daemon teardown, and entries count as fresh for 30 minutes (AGENT_STATUS_STALE_AFTER_MS).
After the daemon restart, those agents' tabs are gone from tabsByWorktree. A click then lands in the guard in WorktreeCardAgents.tsx (handleActivateAgentTab, ~line 211): tab missing but agentStatusByPaneKey[paneKey].worktreeId matches → plain return, no activation, no dismissal, no feedback.
Per the comment, the guard keeps rows alive whose tab "hasn't hydrated yet". That's right for an orchestration worker mid-spawn, but after a daemon restart the tab never arrives, so the row is permanently inert. WorktreeCardAgents.activation.test.tsx ("keeps a live worktree-attributed row visible while its tab is hydrating") pins the silent no-op for the transient case; the never-hydrating case has no exit.
Repro:
Open a remote SSH project and start an agent in a worktree.
On the host, kill the relay daemon (pkill -f relay.js) and let orca reconnect.
Click the agent's row in the sidebar — nothing happens.
Upd.
Reloading the window (View → Reload) clears the rows and they don't come back, so the stale entries appear to live only in renderer state — main's status snapshot no longer contains those sessions.
Activity
fix(agent-status): clear worktree-attributed orphans on relay/daemon …
fix(agent-status): clear worktree-attributed orphans on relay/daemon …
fix(agent-status): clear worktree-attributed orphans on relay/daemon …