Using any MCP with Codex CLI leaves many orphaned Node processes that peg CPU

What version of Codex is running?

0.53.0

What subscription do you have?

Pro

Which model were you using?

gpt-5-high

What platform is your computer?

Darwin 25.1.0 arm64 arm

What issue are you seeing?

After invoking any MCP server from Codex CLI, multiple background node processes are spawned (watchers/helpers). When I quit/close Codex, these node processes keep running and consume significant CPU. CPU usage returns to normal only after I manually kill those node processes in Activity Monitor or via the terminal.

What steps can reproduce the bug?

1. Start Codex CLI with few MCP servers enabled.
2. Use any MCP tool (e.g., run a simple request).
3. Open Activity Monitor (or run ps/pgrep) and observe many “node” processes still running with high %CPU.

What is the expected behavior?

When Codex disconnects from MCP, all child MCP server processes (and their workers) should be terminated cleanly. No lingering node processes and no ongoing CPU usage.

Actual behavior
Dozens of node processes remain alive after Codex exits, often with PPID=1 (re-parented to launchd). They continue consuming CPU until they are manually killed. Once killed, overall CPU usage immediately returns to normal

Additional information

_No response_