App is almost unresponsive for 5 mins after starting up

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

• #15963

Powered by Codex Action

After it starts, please use /feedback to upload your logs and paste the thread ID here.

i'm not sure if it is a duplicate as my unresponsive time is certainly more than 30s

<img width="1422" height="1558" alt="Image" src="https://github.com/user-attachments/assets/cfc0c87b-6830-424a-9892-7fc15bcd3105" />

I keep getting this error when submitting feedback from the codex app @etraut-openai what should I do to generate a thread id?

Ok this is the session ID I got from the conversation by right clicking and "copy session ID"

019d383b-6ed1-73d3-a677-0fb074990837

Let me know if I need to provide any other stuff @etraut-openai

Same here on Windows as well.

Edit: Seems kinda related to my first issue here https://github.com/openai/codex/issues/13589#issuecomment-4006912394
I now disabled all kinds of notifications/popups from Codex app and so far no freezing.

is my thread id accessible and provide useful info? any follow up i need to do?

Please have the official release the new loader and thread manager as soon as possible. Every time it starts, it takes 5 minutes to wait for the initialization to complete, and after loading, the historical chat records of the threads still take 2 minutes to load.
My friend advised me: You better not close VSCode to avoid wasting time.
Rust is safe, but its startup time is no less than booting a computer.

this has been a week... anybody looking into this? as codex app is not open sourced, i don't think the community can do anything about this

Same here. Takes more than 5 minutes to become responsive. Have to switch back to vscode. Windows app does not work at all.

Subject: Codex App for macOS: local app-server blocks UI RPCs for 80-950s under heavy real-world session history

Environment:

• macOS, Codex App for macOS
• App version observed in logs: 26.422.62136
• Electron: 41.2.0
• Codex CLI/app-server reported version: 0.126.0-alpha.8
• Machine: Apple Silicon M5 Max, 128 GB RAM
• Network to api.openai.com appears healthy: unauthenticated /v1/models checks returned in ~0.23-0.30s total

Issue:
Codex App becomes extremely slow on startup and during normal navigation:

• long hang on startup/loading screen
• very slow switching between sessions
• model list takes a long time to appear
• submitting/steering a turn can time out

This does not look like an OpenAI API/network latency issue. The local Electron UI -> Codex app-server RPC path is returning very slowly.

Evidence from Codex App logs:
Current log file:
~/Library/Logs/com.openai.codex/2026/04/28/codex-desktop-4d8304d2-d28d-476b-959e-fb4cc961f49c-72130-t0-i1-141200-0.log

Startup:

• app-server transport started at 14:12:03
• initialize handshake succeeded in 4048 ms
• renderer routes mounted only after 34212 ms
• getAuthStatus timed out after 30000 ms

Representative slow RPCs:

• config/read: 956583 ms
• experimentalFeature/enablement/set: 926487 ms
• thread/resume: 354896 ms
• account/read: 354861 ms
• plugin/list: 354594 ms
• thread/list: 151375 ms
• account/read: 150006 ms
• config/read: 146715 ms
• experimentalFeature/list: 98698 ms
• model/list: 14253 ms
• app/list: 15234 ms
• turn/steer timed out after 30000 ms, later routed after 91719 ms

Process evidence:

• /Applications/Codex.app/Contents/Resources/codex app-server --analytics-default-enabled
• app-server CPU observed between ~74% and ~157%
• app-server RSS observed between ~3.3 GB and ~6.6 GB
• Codex renderer CPU observed up to ~114%

Local state/workload:

• ~/.codex is ~41 GB
• ~/.codex/sessions is ~39 GB
• 1179 rollout JSONL files
• 75 session JSONL files >100 MB
• 21 session JSONL files >500 MB
• 7 session JSONL files >1 GB
• logs_2.sqlite around ~235-258 MB
• state_5.sqlite has 951 threads
• Some thread titles / first_user_message fields are very large, up to ~122 KB

logs_2.sqlite showed active high-volume logging around responses_websocket / tokio-tungstenite / SSE while UI RPCs were delayed.

Hypothesis:
This appears to be head-of-line blocking or task starvation inside the local Codex app-server. Heavy live session processing, JSONL/session writes, SQLite/log writes, plugin/config/account/thread/model UI RPCs appear to share a bottleneck. Lightweight UI operations like thread/list, config/read, account/read, and model/list should not wait tens or hundreds of seconds behind live session processing.

Expected behavior:

• thread/list, config/read, account/read, model/list should remain responsive even when a long active session is streaming or writing rollout data.
• model/list should not take 14s locally when network to api.openai.com is ~0.25s.
• UI navigation and startup should degrade gracefully with large session history.

Request:
Please investigate app-server scheduling/locking around:

• thread/list and session index loading
• model/list path
• config/account/plugin list RPC handling
• SQLite access/log writing contention
• large rollout JSONL files
• high-volume TRACE/WebSocket/SSE logging
• whether active session processing can block unrelated UI RPCs

I can provide logs and sanitized command output if helpful.

Windows 11, same issue. I click the icon to start, I wait for 5 sometimes 15m until the window just appears. Now after a few sessions it starts and completely crashes. I uninstalled, reinstalled but that fixed nothing. I am a new user - I use it around a week or two. Never had issues with Cursor or VS Code. I have 32GB Ram.