sandbox escalation approval times out after Codex update; CLI cannot access macOS Keychain

Summary
After updating Codex Desktop, commands that require sandbox escalation no longer run. The escalation reviewer times out before approving the command. As a result, tools such as lark-cli cannot access macOS Keychain from Codex, even though the same commands work successfully in the user's normal terminal.

Impact
This blocks Codex from using locally configured CLI tools that store credentials in macOS Keychain. In this case, Feishu/Lark CLI is installed and authenticated, but Codex cannot read documents or write Base records because it cannot run lark-cli outside the sandbox.

Environment
Codex app binary: /Applications/Codex.app/Contents/Resources/codex
Codex CLI version: codex-cli 0.133.0-alpha.1
macOS: 26.5, build 25F71
Architecture: arm64
Workspace: /Users/roger/Desktop/AI Product OS
Filesystem sandbox mode: workspace-write
Feishu/Lark CLI path: /Users/roger/.npm-global/bin/lark-cli
Feishu/Lark CLI version: 1.0.39
Expected Behavior
When Codex requests escalated execution for a command such as:

lark-cli auth status
the approval flow should complete, the command should run outside the sandbox, and the CLI should be able to read the macOS Keychain credential.

Actual Behavior
Escalated commands fail before execution with:

Rejected("The automatic permission approval review did not finish before its deadline. Do not assume the action is unsafe based on the timeout alone. You may retry once, or ask the user for guidance or explicit approval.")
When the same command is run without escalation, Codex's sandbox cannot access Keychain and returns:

{
"ok": false,
"error": {
"type": "config",
"message": "keychain Get failed: keychain not initialized",
"hint": "The keychain master key may have been cleaned up or deleted. If running inside a sandbox or CI environment, please ensure the process has the necessary permissions to access the keychain, you can try running this outside the sandbox. Otherwise, please reconfigure the CLI by running lark-cli config init."
}
}
Important Control Check
The same commands work in the user's normal macOS terminal:

lark-cli auth status
lark-cli doctor
The user reports lark-cli doctor returns ok: true in the terminal, including valid bot/user identity and reachable Feishu endpoints. This indicates the Feishu CLI, user authorization, and macOS Keychain are healthy outside Codex.

Reproduction Steps
Install/configure Feishu CLI:
npx @larksuite/cli@latest install
lark-cli config init --new
lark-cli auth login --recommend
Confirm in normal terminal:
lark-cli doctor
It returns ok: true.

In Codex Desktop, run:
lark-cli auth status
Without escalation, it fails with keychain Get failed.

In Codex Desktop, request escalated execution for:
lark-cli auth status
The escalation attempt times out with the approval reviewer deadline error and the command does not run.

Regression Signal
The same workflow worked before the Codex Desktop update. Earlier in the same project, Codex successfully executed Feishu CLI commands with sandbox escalation, read Feishu documents, and wrote to Feishu Base. After the Codex update, all lark-cli commands requiring Keychain access are blocked by escalation approval timeout.

Requested Help
Please investigate whether Codex Desktop 0.133.0-alpha.1 has a regression in sandbox escalation approval handling on macOS, especially for commands that require Keychain access. A workaround or instructions for resetting the local approval reviewer would also help.