Repeated 500 Errors on 5.2 Codex

[krubenok]

Describe the bug

Fairly long conversation and will end up with errors like this with GPT 5.2 Codex, have also seen it with plain 5.2 as well as 5.1 Codex:

✓ Read docs/modernization-ledger.md lines 90-140
   └ 51 lines read

 ✓ Create .devcontainer/devcontainer.json

 ✗ Model call failed: "500 500 Internal Server Error\n" (Request ID:
   61AC:17FDB7:22A6A:2C07A:6969773C)

 ✗ Model call failed: "500 500 Internal Server Error\n" (Request ID:
   61AC:17FDB7:31DEE:3F3F1:696977B0)

 ✗ Model call failed: "500 500 Internal Server Error\n" (Request ID:
   61AC:17FDB7:41B87:53280:69697828)

 ✗ Model call failed: "500 500 Internal Server Error\n" (Request ID:
   61AC:17FDB7:5118A:6682C:6969789F)

 ✗ Model call failed: "500 500 Internal Server Error\n" (Request ID:
   61AC:17FDB7:63BD4:7DBBA:69697928)

 ✗ Model call failed: "500 500 Internal Server Error\n" (Request ID:
   61AC:17FDB7:74CEE:92EA7:696979A2)

 ✗ Execution failed: Error: Failed to get response from the AI model; retried 5 times (total retry
    wait time: 84.32230497947954 seconds)

Affected version

0.0.382 Commit: 18bf0ae

Steps to reproduce the behavior

Long (ish) running conversation with an OpenAI Model.

... somewhat randomly ...

500 errors.

Expected behavior

No 500 errors - either compaction, rate limits or something explainable.

Additional context

macOS, running in Ghostty. Not an issue with Claude models only GPT 5.x models.