Remote Computer tasks take 12–21 seconds to load and repeatedly reconnect

Codex Bugs
,
1

Remote Computer tasks take 12–21 seconds to load and repeatedly reconnect

Environment

  • Surface: Codex desktop App
  • Codex version: ChatGPT/Codex App 26.915.31945 (build 9922); remote app-server 0.155.0-alpha.9.2
  • OS and IDE/browser, if relevant: macOS 26.6.2 (25G83), Apple silicon (arm64)
  • Model and plan or API-key authentication: GPT-5.6 Sol, Medium reasoning; ChatGPT account authentication, not API-key authentication

Bug

Codex Remote Computer tasks hosted on my Mac Pro are extremely slow to open. A cold task load takes approximately 12–21 seconds and can sometimes remain blank or display “Loading task.”

The Mac Pro itself is not resource-constrained: it has approximately 36 GB of available memory, no swap pressure, 1.5 TB of free disk space, and 64–77% CPU idle.

I expected remote tasks to open within a few seconds, as they previously did and as my direct SSH connections generally do.

The problem appears to be in Codex’s Remote Computer WebSocket and task-hydration path. The connection repeatedly becomes unknown, reconnects, and reloads task history using several sequential requests.

Reproduction

  1. Connect a Mac as a Codex Remote Computer.
  2. Create or open a Codex task hosted on that Mac.
  3. Leave the task idle so that reopening it requires a cold load.
  4. Open the task from the Codex desktop sidebar.
  5. Observe the blank/loading state for approximately 12–21 seconds.
  6. Reopen the same task immediately. It may load quickly while warm.
  7. Allow the Remote Computer connection to reconnect and repeat the test. The long cold-loading delay returns.

This occurs across multiple Mac Pro tasks, not just one conversation.

Diagnostics

Observed timings:

  • Cold task hydration: 21.1 seconds
  • Another cold task open: approximately 12 seconds
  • Older history-page request: 13.1 seconds
  • Warm reopen of the same task: 0.29 seconds
  • Individual Remote Computer requests commonly take approximately 1.25–1.35 seconds
  • Equivalent local task operations generally complete in 1–270 ms
  • Creating a new task on the Remote Computer took approximately 30 seconds

Relevant sanitized log messages:

remote control app-server stream became unknown

thread not loaded: <thread-id>

No AppServerManager registered for hostId: remote-control:<redacted>

App server request expired while queued

Codex app-server is not available

websocket_reconnect_recovery_start
websocket_reconnect_marked_threads_needing_resume

The affected Remote Computer app-server currently reports:

0.155.0-alpha.9.2

It previously reported:

0.154.0-alpha.6.2

The local task database is responsive, and indexed history queries complete essentially instantly. Removing an offline Remote Computer stopped its retry traffic but did not resolve the Mac Pro loading delay.

When it happened and how often:

  • Investigated on September 20, 2026, approximately 16:15–17:00 (UTC+2).
  • Reproduces consistently when opening a cold Mac Pro task.
  • Warm tasks can load quickly until the Remote Computer connection or task state is rehydrated.

A screenshot of the blank task-loading state is attached.

2

You posted this in the appropriate place on the forum for discussing and documenting a Codex bug. The forum, however, is not the official issue tracker. If the problem needs to be formally reported and tracked by the Codex maintainers, the official place is the OpenAI Codex GitHub issue tracker.

Thanks for following the posting requirements for the Codex Bugs subcategory:

I asked ChatGPT to search for the closest related issue, and it identified:

https://github.com/openai/codex/issues/32771

If that issue describes the same problem you are experiencing, consider adding a :+1: reaction to the GitHub issue itself, rather than only reacting to this forum topic. GitHub reactions can help the maintainers gauge how many users are affected.

If it does not match your problem closely enough, search the existing Codex issues for a better match. If you cannot find one, consider opening a new issue and include as much relevant information as possible, such as your Codex version, operating system, environment, reproduction steps, expected behavior, actual behavior, and any relevant logs or error messages.