Account-specific `server_is_overloaded` on GPT-5.6 Sol: one Plus account fails while two other Plus accounts work on same machine/network

Subscription

ChatGPT Plus

Issue

One of my ChatGPT Plus accounts consistently fails in Codex when using GPT-5.6 Sol with the following server-side error:

{
  "error": {
    "type": "service_unavailable_error",
    "code": "server_is_overloaded",
    "message": "Our servers are currently overloaded. Please try again later.",
    "param": null
  },
  "sequence_number": 2
}

The Codex UI may also show:

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

Account comparison / isolation test

I have three ChatGPT Plus accounts and tested them under the same environment:

• Same computer
• Same network
• Same Codex client/environment
• Same GPT-5.6 Sol model

Results:

• Plus account A: repeatedly returns service_unavailable_error / server_is_overloaded
• Plus account B: works normally
• Plus account C: works normally

Because the two other Plus accounts work on the same machine and network, this appears to be account-specific rather than a local network, computer, or general model-capacity issue.

Steps to reproduce

1. Open Codex.
2. Sign in with the affected ChatGPT Plus account.
3. Select GPT-5.6 Sol.
4. Send a normal coding request.
5. The request repeatedly fails with server_is_overloaded / Selected model is at capacity.
6. Sign out and sign in with either of two other ChatGPT Plus accounts on the same computer and network.
7. Repeat the same request using GPT-5.6 Sol.
8. The other accounts work normally.

Expected behavior

The affected Plus account should be able to use GPT-5.6 Sol normally when capacity is available, consistent with the other Plus accounts in the same environment.

Actual behavior

The affected account repeatedly receives a server-side overload error while the other two Plus accounts work normally under the same conditions.

Possible area to investigate

Could the team please check whether this account has an issue with any of the following?

• account-specific backend/model routing
• account provisioning / entitlement state
• admission or capacity allocation
• account-level throttling
• quota / rate-limit state

This looks similar to #44113, but in this case I have two separate Plus accounts working normally as controls while one account consistently fails.

Thank you.

Start with the account comparison and reproduction steps in the issue, using the same machine, network, Codex environment, and GPT-5.6 Sol model. Investigate the affected account's backend routing, entitlement, admission, throttling, and quota state; done means identifying and resolving the account-specific cause or documenting why the account cannot access the model.