Switching to sub-agent & back to main agent via `/agent` repeats `request_user_input` elicitation

What version of Codex CLI is running?

v0.104.0

What subscription do you have?

Enterprise

Which model were you using?

gpt-5.3-codex

What platform is your computer?

Linux 4.18.0 ... x86_64

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

_No response_

What issue are you seeing?

After using /plan mode with a request_user_input call, switching to a sub-agent (/agent) and back again repeats the request_user_input questioning/elicitation again.

Uploaded feedback for thread 019c8b3f-6a46-7fa2-9404-c1005f462c53.

What steps can reproduce the bug?

1. Working in /plan mode.
2. Agent uses request_user_input; submit answers.
3. Agent turn ends, switch back to default collab mode.
4. Ask the agent to implement the plan.
5. Agent spawns a subagent.
6. /agent to switch to subagent.
7. /agent to switch back.
8. Previous questions from request_user_input are immediately repeated.

What is the expected behavior?

Don't repeat the questions/elicitation!

Additional information

_No response_