[Bug] Intermittent "Selected model is at capacity" errors in Codex Desktop (all models) on ChatGPT Pro

Describe the bug
On a ChatGPT Pro subscription, Codex Desktop keeps throwing this error
intermittently (works for a while, then fails, then recovers):

Selected model is at capacity. Please try a different model.

It happens frequently and interrupts long-running tasks, which do not resume
on their own. Importantly, it occurs with any model — not a single
high-demand one — which suggests this is not a per-model capacity issue.

Environment

• Plan: ChatGPT Pro
• Client: Codex Desktop
• Model: any / all models (reproduced across multiple models)
• OS: macOS 26.5 (25F71)

To Reproduce

1. Open Codex Desktop and select any model.
2. Start a task / continue working.
3. After some time, the next request fails with the capacity error,
   regardless of which model is selected.

Expected behavior
A Pro subscriber should not hit model-capacity limits, or the task should
retry / recover gracefully.

Actual behavior
The request fails with "Selected model is at capacity. Please try a different
model." across all models, and the task stops without automatic recovery.

Frequency
Intermittent but frequent — occurring multiple times per day since
September 7, 2026, recurring after temporary recovery.

Additional context
Similar to existing reports #41790 and the "GPT-5.6 Sol repeatedly hits
'Selected model is at capacity' in Codex Desktop" thread on the OpenAI
developer community.

Start by reproducing the intermittent capacity error in Codex Desktop on macOS with several models during a long-running task. Review reports #41790 and the linked developer-community thread for comparison, then trace whether the failure is client-side or service-side. Done means requests no longer fail across models, or the task retries and recovers automatically.