VS Code Codex chat webview turns gray while app-server continues running

Description

The Codex chat panel in VS Code intermittently turns completely gray during a conversation. The Codex backend continues running and produces reasoning-summary events, but the chat UI becomes unusable.

This happens on the latest Codex extension version.

Environment

• OS: macOS 26.5.2 (Build 25F84)
• Architecture: Apple Silicon / arm64
• VS Code: 1.130.0
• VS Code commit: 1b6a188127eeaf9194f945eb6eb89a657e93c54c
• Codex VS Code extension: openai.chatgpt@26.721.41059
• Connection: Local VS Code window, not Remote-SSH/WSL
• Subscription: [Plus]

Steps to reproduce

1. Open a Codex chat in VS Code.
2. Run a relatively long task or conversation.
3. Leave the chat running for several minutes.
4. The Codex panel eventually becomes gray and stops responding.

The issue is intermittent. The issue happened when I tried to submit a message when the agent was running.

Expected behavior

The Codex chat UI remains visible and interactive while the task continues.

Actual behavior

The chat webview becomes gray. The backend appears to continue working in the background.

Relevant log excerpts

The extension logs repeatedly contain:

[IpcClient] Received broadcast but no handler is configured method=client-status-changed

Other unhandled IPC events include:

• query-cache-invalidate
• thread-stream-following-changed
• thread-queued-followups-changed
• thread-read-state-changed

Shortly before the UI became unusable, the log reported:

[desktop-notifications][global-error] ResizeObserver loop completed with undelivered notifications.

The app-server continued emitting successful reasoning-summary events after this.

The extension was later activated again and resumed the conversation.

Additional observations

Authentication succeeds, and conversations continue processing. This does not appear to be an authentication or model request failure.

There are also unrelated-looking warnings about stale/unsupported feature keys:

• thread_tools
• apps_mcp_path_override

and a plugin manifest warning for a defaultPrompt longer than 128 characters.

Diagnostic data

codex-log-sanitized.txt

The issue occurred during a long-running conversation on 2026-07-27. The Codex UI became gray, but the app-server continued producing successful reasoning-summary events afterward. The issue happened when I tried to submit a message when the agent was running.

The extension was activated at:

• 13:50:01
• 14:00:11

Relevant warnings/errors:

[IpcClient] Received broadcast but no handler is configured method=client-status-changed
[IpcClient] Received broadcast but no handler is configured method=query-cache-invalidate
[IpcClient] Received broadcast but no handler is configured method=thread-stream-following-changed
[IpcClient] Received broadcast but no handler is configured method=thread-queued-followups-changed
[IpcClient] Received broadcast but no handler is configured method=thread-read-state-changed

Shortly before the UI became unusable:

[desktop-notifications][global-error]
ResizeObserver loop completed with undelivered notifications.

The app-server continued emitting events such as:

Reasoning summary item completed

Additional state-related warning:

codex_rollout::list: state db discrepancy during find_thread_path_by_id_str_in_subdir: falling_back

Authentication succeeded and the conversation continued processing, so this does not appear to be a model request or authentication failure.

A sanitized full log is attached: codex-log-sanitized.txt.