Codex Team: Please address Sub-Agent system - it feels brittle and unreliable

[Allmight97]

Here is an issue I created with Codex in response to yet another session with sub-agent friction that represents a very common failure pattern using sub-agents.

e.g. Codex main agent invokes agents by itself or upon request, and then it either interrupts them when they haven't responded in X amount of time, or it says they give nothing useful or nothing at all, and sometimes they hang, and it has to intervene, and the main codex agent does the work itself anyway.

This adds extra churn and it really feels like the sub-agent system as it currently exists is brittle. That combined with the increasing performance issues with the app when folks have many chats that are not archived yet and are using the memory system leaves me a little concerned that we're getting ahead of ourselves with adding new features and abilities when mainline features like Sub-agents need more attention.

If I were testing this for my team, this would be a major source of apprehension when deciding to adopt codex App for team work and the like.