Write tool with large content causes exec-server relay disconnect → reconnect loop since 0.140+

What version of Codex CLI is running?

0.141.0

What subscription do you have?

pro20x

Which model were you using?

gpt-5.5-codex

What platform is your computer?

windows

What terminal emulator and version are you using (if applicable)?

_No response_

Codex doctor report

What issue are you seeing?

## Environment

• Codex CLI version: 0.141.0 (also reproducible on 0.140.x)
• OS: Windows 11 Pro 10.0.26200
• Shell: Git Bash (MINGW64)
• Terminal: Windows Terminal

## Expected Behavior
File writes successfully.

## Actual Behavior

• Short files (under ~200 lines): works fine.
• Long files: Codex begins writing, then shows "Reconnecting", reconnects, retries the write, disconnects again →

infinite loop. The file never completes.

• Downgrading to 0.139.0 completely resolves the issue. Upgrading back to 0.140+ reproduces it consistently.

## Likely Related Changes

• #28286 (restore exec-server relay keepalives) — suggests keepalive handling changed around the threshold that breaks

large tool-call results

• #28234 (MCP tool timeout increased to 300s) — suggests timeout values were recently adjusted
• The large session memory/latency optimizations in 0.140/0.141 may have introduced a regression where large Write

tool call payloads exceed the relay timeout or trigger a premature keepalive disconnect

## Workaround
Downgrade to 0.139.0: npm install -g @openai/codex@0.139.0

What steps can reproduce the bug?

1. Install latest Codex CLI (npm install -g @openai/codex)
2. In any directory, start a codex session
3. Ask it to write a long file (e.g., a 500+ line markdown document or a 300+ line code file)

What is the expected behavior?

_No response_

Additional information

_No response_