[Bug]: Subagent failure notification keeps a red Failed badge while the resumed subagent runs #17486

Before submitting

• I searched existing issues and did not find a duplicate.
• I included enough detail to reproduce or investigate the problem.

Area

apps/web

Steps to reproduce

1. In a Claude thread (claudeAgent), start several background subagents
2. Let them fail on the five-hour usage limit. Each failure posts a Subagent "…" failed notification row in the thread
3. After the reset, resume each subagent with SendMessage to the same agent ID (here the parent did this after a "continue" message)
4. Watch those notification rows while the subagents work

Expected behavior

The row shows the subagent's live state (Running, then its final outcome), or it shows the failure clearly as a past event and does not mix it with live progress

Actual behavior

The badge stays a red Failed while the progress line under it updates live, for example "Running Run broader REST-related test files". The server already has the right state: all four subagent items are stored with status: "running", updated between 12:48 and 12:49 UTC

From the installed _chat bundle, the notification row builds an event from the notification's outcome ({ failed: "Failed", … }) and prefers it over the live status:

o = i?.status ?? e.status          // live subagent status
s = e.event ? e.event.status : o   // badge status
c = e.event?.label ?? tI(o).label  // badge label

The progress text comes from the live subagent, so the row combines an old outcome with current progress

Related reports

• [Bug]: Subagent status stays red after a usage limit recovery, and never updates while running #7314 and [Bug]: A Claude subagent resumed with SendMessage stays "completed" or "failed" in the Agents panel (regression of #5529) #14002 covered this in V1 and were closed when Orchestrator V2 (feat(orchestrator): introduce new orchestrator #2829) merged, with a request to open a new issue if it persists on a later nightly. It does. The V1 fix (fix(client-runtime): a resumed Claude subagent no longer stays failed or stopped in the Agents panel #14696) was closed unmerged
• [Bug]: Usage-limited Claude thread is never armed for auto-resume when background-task wake runs finish as completed right after the limit stop #16815: The same incident. These failure notifications are the wake runs that kept the thread from auto-resuming

Possible fix

Let a later start of the same subagent override the notification's outcome, for example when the subagent's startedAt is after the notification's createdAt. Or label the outcome as past ("Failed at 14:21") and keep it apart from the live status

Impact

Minor bug or occasional failure

Version or commit

T3 Code (Nightly) 0.0.46-nightly.20261009.2861 (3b6af0bd1600), Claude Code CLI 2.1.295

Environment

Windows 10 Pro 10.0.19045, desktop app with its local server, provider claudeAgent, model claude-opus-5-5

Logs or stack traces

Screenshots, recordings, or supporting files

image.png

Workaround

No response