Stream Disconnected Before Completion (max_output_tokens)

Description
When submitting requests via Codex CLI with the GPT-5.3-Codex model, I repeatedly encounter the following error:

stream disconnected before completion: Incomplete response returned, reason: max_output_tokens

The response is consistently truncated, and the request fails to complete successfully.

Background

• I am using the exact same GPT-5.3-Codex model and identical usage methods/CLI commands as I have in the past.
• This error never occurred previously—it started appearing suddenly with no changes to my local configuration, request logic, or input parameters.
• Thread ID for reference: 019cf20e-b58d-7e62-a5ae-d658b3a801aa

Key Questions

1. Could this issue be attributed to an underlying problem with the GPT-5.3-Codex model (e.g., recent backend updates, token limit adjustments, service degradation, or API behavior changes)?
2. What is the correct/recommended value/range for max_output_tokens (or max_tokens—since Codex CLI uses max_tokens in configuration) to resolve this incomplete response error? Please specify the optimal setting for general code-related use cases (e.g., code generation, debugging, small-to-medium module development).

Additional Context

• The requests are standard code-focused tasks (no extremely long output requirements) that worked reliably with default settings before this issue arose.