openai / openai/codex

[Windows] Reconnecting 5/5 and severe slowdown with 4 concurrent Luna Max tasks

Open
#37,938 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

app bug connectivity performance windows-os
Dominant language
Rust
Stars
125k
Forks
19.5k
PR merge metrics
PR metrics pending

Description

Hi Codex team,

Starting today, Codex on Windows has become extremely slow. I normally run four project tasks concurrently, mostly using GPT-5.6 Luna with Max reasoning.

This workflow worked normally before, but now the tasks frequently stall, remain in “Thinking” for a long time, and display “Reconnecting 5/5.”

Expected behavior

I need four project tasks to run concurrently and complete reliably. Reducing the number of concurrent tasks is only a temporary workaround and does not fit my normal workflow.

Actual behavior

  • Frequent “Reconnecting 5/5”
  • Tasks remain in “Thinking” for a long time
  • Projects take much longer than they did previously
  • The issue repeatedly occurs when several tasks are active

Environment

  • Windows Codex desktop app
  • App version: 26.803.10989.0
  • Model: mostly GPT-5.6 Luna
  • Reasoning level: Max
  • Concurrent workload: 4 project tasks

Request

Could you please investigate whether this is a client or concurrency regression?

Because the stalled and reconnecting runs have consumed usage without completing normally, I would also really appreciate a usage reset if possible.

Thank you, for your help. I really appreciate it, and thank you to the entire Codex team for looking into this.

This is my account name :@euanchow

My Codex account ID is:2730facb-2f37-44db-aec9-69a63161767f

And can you give me a reset? Thank you !!!!
Image

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the report in the Windows Codex desktop app version 26.803.10989.0 with four concurrent GPT-5.6 Luna tasks at Max reasoning. Observe the “Reconnecting 5/5” and prolonged “Thinking” behavior, then determine whether the regression is client-side or concurrency-related. Done means the four-task workflow completes reliably without the reported stalls or severe slowdown.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust
Domain
desktop, networking, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.