omarchy-agent-usage-codex: account/rateLimits/read RPC timeout (4s) misreported as unavailable #12880

What happened

The Codex agent-usage collector sometimes reports no rate-limit data. The panel and a custom alert script show "Codex limits unavailable." The stored authHelpText is the literal string account/rateLimits/read — this is the exception text from a timeout, not an auth message.

Root cause

fetch_codex_rpc() in bin/omarchy-agent-usage-codex calls rpc_request(proc, 3, "account/rateLimits/read", timeout=4). The Codex app-server process starts fresh on every collector run. Under normal dev load (other Codex or Claude Code sessions, background builds), this call sometimes exceeds 4 seconds. rpc_request() raises TimeoutError(method). The collector stores str(exc) as authHelpText, which prints only the method name, not a timeout description.

What I expected

Either the RPC call finishes inside its timeout, or the failure text reads as a timeout, not an auth problem.

Steps to reproduce

1. Run several concurrent Codex or Claude Code sessions, or another CPU-heavy job.
2. Run omarchy agent usage-update codex --force a few times, or wait for the panel's normal refresh.
3. Watch ~/.local/state/omarchy/agents/usage/codex.json. Under load, usageStatusText becomes "Codex limits unavailable" and authHelpText becomes "account/rateLimits/read".
4. The next refresh, with no code change, often succeeds. This confirms a timing race, not an auth failure.

Suggested fix

1. Raise the timeout on the account/rateLimits/read call, or retry once before giving up.
2. Give a timed-out rpc_request() call a clear message, so authHelpText does not read like a permission problem.

System

• Omarchy 4.0.4-1, Hyprland 0.56.2
• Codex CLI 0.155.1, plan: plus
• Framework Laptop 13 Pro (AMD Ryzen AI 300 series)