[BUG] "Claude Code on <host> needs to sign in again" doesn't say which Claude Code instance, or that a CLI login is required

Environment: macOS 26 (Darwin 25.6.0), Claude Code CLI 2.1.259 via nvm, desktop app (Code tab), mobile app connected through claude remote-control under launchd.

What happened

The mobile app repeatedly showed: "Claude Code on needs to sign in again. Sign in there, then send your message again."

Signing in on the Mac didn't clear it. The message doesn't say which Claude Code needs the login: the desktop app, a terminal session, or a background claude remote-control process. Signing into the desktop app doesn't refresh the CLI's keychain credential, so the banner persisted until the CLI itself was re-authenticated with claude auth login.

Expected

• Name the instance (e.g. the --name given to remote-control, or "CLI" vs "desktop app").
• State the actual remedy: run claude auth login on that machine (or /login in a terminal session).
• Ideally, have a long-running remote-control process surface the auth failure in its own log/status line ("auth expired — run claude auth login") instead of only showing Reconnecting.

Repro

1. Run claude remote-control --name X under launchd; keep it up for weeks.
2. Let the OAuth token expire.
3. Send a message from the mobile app → banner appears.
4. Sign into the desktop app → banner remains. claude auth status in a terminal reports loggedIn: false.
5. claude auth login → banner clears.

Start by reproducing the flow with claude remote-control --name X, an expired token, and the mobile app banner; compare claude auth status, claude auth login, and /login. Trace how the remote-control instance identity and authentication failure reach the banner and process status. Done means the affected instance is identified, the CLI login remedy is explicit, and long-running remote-control reports the auth failure instead of only Reconnecting.