Work mode: exec-server initialization times out after 10s, leaving file editing unavailable

What version of the Codex App are you using?

Exact client and bundled runtime versions were not captured during the failure. Observed in ChatGPT Work mode on 2026-09-14.

What subscription do you have?

Omitted from this public report for privacy.

What platform is your computer?

The affected surface is a Work-mode task with a managed workspace. Client OS and runtime build are not independently verified here; this report does not attribute the failure to a particular local OS.

What issue are you seeing?

A workspace connection failure leaves the task able to continue conversational work and use connected read tools, but unable to edit and save the requested file.

The environment diagnostic reported status failed with this error:

exec-server connection attempt failed: timed out waiting for exec-server initialize handshake after 10s

During the affected turn, workspace shell/file-editing capabilities were unavailable in the discovered tool set. Connected file reads still succeeded. The assistant completed the requested content in its reply but could not apply it to the existing document. Reading a file through a connector is not evidence that the workspace filesystem itself was reachable.

The user reports this as recurring. One exact diagnostic occurrence is documented here; the total frequency and root cause have not been measured.

What steps can reproduce the bug?

Observed workflow (intermittent; not a deterministic standalone reproducer):

1. Open or resume a Work-mode conversation involving an existing document.
2. Ask the assistant to update that document and save the result.
3. The managed workspace fails to initialize with the 10-second exec-server handshake timeout above.
4. Conversation and some connected reads remain available, but the file update cannot be completed in that turn.

A generic prompt for testing this workflow is: “Add one paragraph to the existing Markdown document and save the updated file.” No original prompt or document content is included.

What is the expected behavior?

• Detect the failed workspace before committing to a file-editing workflow.
• Attempt a bounded reconnect or offer a clear workspace recovery action that preserves the conversation and existing files.
• Refresh available execution/file-editing capabilities after successful recovery.
• Resume the pending edit and verify that the saved artifact contains it.
• If recovery fails, distinguish drafted content from saved changes and provide an actionable recovery status.

Additional information

This blocks a core deliverable: an updated file. Returning the text in chat does not complete that task, and recurring failures force the user to repeat requests or manually transfer edits.

Existing reports were searched using the exact diagnostic and related exec-server/workspace terms. Related but not established duplicates:

• #40400 concerns Work-mode app-server restarts and missing tool outputs on Windows.
• #29524 concerns large writes causing relay disconnect/reconnect loops.

Neither establishes the cause of the initialization failure reported here. Please route this to the managed Work workspace/runtime owner if this repository is not the correct component tracker.

Privacy: no document content, project names, filenames, local paths, workspace/session identifiers, credentials, screenshots, or raw transcripts are attached. No private session sharing was initiated.

Start by tracing the managed workspace exec-server initialization and the 10-second handshake timeout, then reproduce the documented Work-mode workflow with an existing Markdown document. Done means the failure is diagnosed and recovery or reconnect restores file-editing capabilities, resumes the pending edit, and clearly distinguishes saved changes from drafted content when recovery fails.