Codex Windows app-server saturates upload bandwidth and causes packet loss

What version of the Codex App are you using (From “About Codex” dialog)?

OpenAI.Codex 26.611.8604.0

What subscription do you have?

pro

What platform is your computer?

Windows 11 x64

What issue are you seeing?

Codex Desktop app-server appears to saturate my upload bandwidth and causes severe packet loss / high latency.

Problematic process

text
codex.exe app-server --analytics-default-enabled

Process path:

text
C:\Program Files\WindowsApps\OpenAI.Codex_26.611.8604.0_x64__2p2nqsd0c76g0\app\resources\codex.exe

Resource Monitor evidence

Windows Resource Monitor showed:

• Process: codex.exe
• PID: 1452
• Send rate: 1,334,503 B/sec
• Approx upload: 10.7 Mbps
• Receive rate: 0 B/sec in that sample
• Remote address: 104.18.32.x / Cloudflare-backed HTTPS

This upload rate is close to my upstream limit and causes network disruption.

Network impact

I ran a 30-minute local monitor that sampled gateway ping, internet ping, DNS, and interface throughput.

Results:

• Samples: 1099
• Gateway loss: 1
• Internet loss: 219
• DNS fail samples: 0
• Max upload: 16.097 Mbps
• Max internet ping: 453 ms
• Longest continuous internet failure: about 102 seconds

The two longest internet-loss windows happened while upload was saturated:

1. 20:17:21 - 20:20:38

• Continuous internet loss: about 96 seconds
• Average upload: 9.82 Mbps
• Max upload: 16.097 Mbps
• Gateway loss during this run: 0

1. 20:32:21 - 20:35:46

• Continuous internet loss: about 102 seconds
• Average upload: 9.85 Mbps
• Max upload: 14.605 Mbps
• Gateway loss during this run: 0

This suggests the local router link stays mostly reachable, but Codex upload saturation causes upstream bufferbloat / internet packet loss.

Local Codex log churn

The local Codex log database is also very large:

• C:\Users\Owner\.codex\logs_2.sqlite = about 783 MB
• C:\Users\Owner\.codex\logs_2.sqlite-wal = about 342 MB

Recent rows include many TRACE logs related to WebSocket/SSE frame-level activity, for example:

• tokio-tungstenite
• WebSocketStream
• received frame
• Received message ack
• codex_api::endpoint::responses_websocket
• codex_api::sse::responses
• codex_app_server::outgoing_message
• target=log
• codex_otel.log_only
• codex_otel.trace_safe

This looks related to existing app-server / persistent logging churn issues.

Possibly related issues

• openai/codex#29799
• openai/codex#29798
• openai/codex#29589
• openai/codex#29588
• openai/codex#29570

Expected behavior

Codex Desktop should not continuously upload at 10+ Mbps or saturate upstream bandwidth during normal use.

Text prompts, normal coding sessions, or analytics should not cause sustained upload saturation that leads to internet packet loss.

Actual behavior

codex.exe app-server uploads at around 10-16 Mbps, causing high latency and packet loss on the whole machine/network.

Workaround

Fully quitting Codex stops the upload saturation. Reopening Codex may reproduce the issue.

I can provide the CSV network monitor logs if needed, but I am not attaching the full Codex SQLite logs because they may contain sensitive session data.

What steps can reproduce the bug?

I do not have a minimal deterministic reproduction yet, but this reproduced during normal Codex Desktop usage on Windows.

Observed steps:

1. Start Codex Desktop on Windows.
2. Leave the Codex Desktop app-server running:

codex.exe app-server --analytics-default-enabled

1. Use Codex normally for coding / assistant work.
2. Monitor Windows Resource Monitor > Network > Network Activity.
3. Observe codex.exe sending around 1,334,503 B/sec (~10.7 Mbps) to a Cloudflare-backed HTTPS remote address.
4. While this upload is happening, run continuous ping monitoring to the local gateway and 1.1.1.1.
5. Observe internet ping spikes / packet loss while the local gateway remains mostly reachable.

In my 30-minute monitor, the two longest internet-loss windows occurred while upload was saturated:

• 20:17:21 - 20:20:38: about 96 seconds of internet loss, average upload 9.82 Mbps, max upload 16.097 Mbps, gateway loss 0.
• 20:32:21 - 20:35:46: about 102 seconds of internet loss, average upload 9.85 Mbps, max upload 14.605 Mbps, gateway loss 0.

Fully quitting Codex stops the upload saturation.

What is the expected behavior?

Codex Desktop should not continuously upload at 10+ Mbps or saturate upstream bandwidth during normal use.

Text prompts, normal coding sessions, analytics, or background app-server activity should not cause sustained upload saturation, high latency, or internet packet loss.

Additional information

Possibly related: openai/codex#29799, #29798, #29589, #29588, #29570. I can provide CSV network monitor logs if needed, but I am not attaching the full SQLite logs because they may contain sensitive session data.

<img width="536" height="233" alt="Image" src="https://github.com/user-attachments/assets/7bf0da25-c684-4cbe-a3fa-d9452db19a38" />