Repository navigation
[Bug]: Codex subagent completion output can interleave with main-agent output #2477
Description
Summary
Codex subagent completion output can interleave mid-token with main-agent assistant output in the visible chat UI.
The persisted messages/events appear to remain separate, but the rendered chat text can splice the two outputs together.
Example
Expected separate text fragments:
Include constraints: read-only, use a specific tool, don’t modify files.
I’m seeing a clear split: short diagnostic nudges during debugging...
Observed rendered text:
Include constraints: read-only, use a specific tool, don’tI modify’m seeing files a. clear split...
Reproduction Shape
- Use T3 Code with Codex.
- Start a main-agent turn that spawns a background/subagent.
- Have the main agent emit commentary while the subagent is completing.
- Observe the visible chat output when both messages arrive close together.
Expected Behavior
Main-agent messages and subagent completion messages should render as separate chat items, or otherwise remain textually isolated.
Actual Behavior
The UI can render a single garbled assistant output where text from the subagent completion and main-agent commentary are interleaved mid-token.
Evidence
In one observed case, local provider logs showed two separate item/completed events about 72 ms apart:
- one for main-agent commentary
- one for subagent final output
The local SQLite projection also contained separate assistant message rows. This suggests the issue is likely in stream/render/event assembly rather than model output.
Impact
This makes Codex subagent workflows hard to trust because final answers can appear corrupted even when the underlying stored events are distinct.
Related Issues
Possibly related to prior stream/rendering issues:
- [Bug]: Raw internal tool-call metadata leaks into visible assistant message content #1147
- Responses duplicate when Stream assistant messages is on #349
- [Bug]: OpenCode provider can duplicate the assistant response / replay a large suffix chunk at completion #2380
- Subagent support as nested threads #538
Activity
paulhenri-l commentedon Sep 3, 2026
I’m seeing the same issue with Codex. Legacy token-by-token streaming is disabled.
I initially thought this might be a Markdown rendering issue, but the garbled text is already present in T3 Code’s SQLite database. The DOM is simply displaying the corrupted stored message.
For the affected response, I found three
thread.message-sentevents using the same assistant message ID. Their text lengths were 7,977, 716, and 513 characters, and the final projected message was 9,206 characters. Importantly, the first 7,977-character event was already corrupted.One example from the stored text:
This may be slightly different from the original reproduction, where the persisted messages appeared to remain separate. In my case, the output seems to become interleaved before it reaches the UI—possibly while Codex main-agent and subagent output are being ingested around the same time.
lastobelus commentedon Sep 4, 2026
Interesting, I have not seen this issue for quite some time even though I sometimes have 100+ subagent runs in a thread.