Unexpected Bengali-like character injection in ko/en output since Codex 5.3

What version of Codex CLI is running?

codex-cli 0.101.0

What subscription do you have?

plus

Which model were you using?

gpt-5.3-codex medium

What platform is your computer?

Darwin 25.2.0 arm64 arm

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

Warp, iTerm, IntelliJ Terminal

What issue are you seeing?

Since Codex 5.3, responses requested in Korean/English sometimes include unexpected non-ko/en characters (Bengali-like text).

request:

SkuEnableStatusConverter.java 에서 setValue를 option.getStatus() 그대로 하고 있는데
  status 값이 01인 경우에만 1 을 넣고 나머진 0으로 설정하게 바꿔줘.

response

테스트 스타일을 맞추기 위해 기존 SKU 컨버터 테스트 형식을 확인했고, 이제 SkuEnableStatusConverterTest를 먼저 "01" -> 1 케이스 하나로 추가해 Red 상태를 만들겠습니다. এরপর 최소 구현으로 Green 만든 뒤 "기타 -> 0" 케이스를 추가하겠습니다.

"এরপর" is Bengali

What steps can reproduce the bug?

1. Run Codex CLI and request code edits/summaries in Korean.
2. Receive assistant response.
3. Occasionally observe non-ko/en characters mixed into text.

What is the expected behavior?

Output should contain only requested languages (ko/en), excluding code/file paths.

Additional information

_No response_