fix(cron): close codex app-server session in teardown to prevent subprocess leak (#62101)

[liuhao1024]

What does this PR do?

Fixes a subprocess leak in cron jobs when using the codex_app_server runtime. Each scheduled cron run creates a new ephemeral agent, which spawns a codex app-server subprocess. The subprocess should be cleaned up when the job finishes, but agent.close() does not include _codex_session cleanup. This caused one leaked codex app-server subprocess per cron run until the gateway is restarted.

The fix adds a close_codex_session() helper in agent/codex_runtime.py (idempotent, safe to call multiple times) and calls it from cron's _teardown_cron_agent() function in the finally block after agent.close().

Related Issue

Fixes #62101

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

• agent/codex_runtime.py: Add close_codex_session() helper function that safely closes the Codex app-server session if it exists
• cron/scheduler.py: Call close_codex_session() in _teardown_cron_agent() after agent.close() to prevent subprocess leaks

How to Test

1. Configure provider: openai-codex with openai_runtime: codex_app_server
2. Create a recurring cron job: hermes cron create "0 8,13,18 * * *" "echo test"
3. Start the gateway: hermes gateway --platform telegram
4. Wait for a few scheduled runs to complete
5. Check for leaked codex app-server processes: ps -axo pid,ppid,lstart,command | grep "codex app-server"
6. Expected result: No codex app-server subprocesses remain after each cron job completes

Observed result: The issue reporter verified locally that after this patch, a manual hermes cron run leaves no codex app-server process behind once the job completes.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 26.5 (arm64)

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

[teknium1]

Thanks for the focused cron cleanup. The premise is confirmed on current main: agent/codex_runtime.py:325 creates a per-agent app-server session, while cron/scheduler.py:3372-3385 tears down cron agents through agent.close() without closing that session.

Problems

• The diff at 7313d61c62a2 changes only agent/codex_runtime.py and cron/scheduler.py; it adds no regression test. Existing teardown-order coverage in tests/cron/test_run_one_job.py:165-205 verifies delivery before agent.close() but does not exercise a Codex session.

Suggested changes

• Add a cron teardown regression that asserts _teardown_cron_agent() closes and clears an attached _codex_session, including when agent.close() raises. This validates the exact lifecycle path changed here.

Automated hermes-sweeper review.