Inline fenced Markdown blocks render inconsistently between Codex Desktop tasks

What version of the Codex App are you using (From “About Codex” dialog)?

26.908.4834.0

What subscription do you have?

ChatGPT Pro

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

A predecessor Codex Desktop task emitted valid inline Markdown with fenced code blocks, but the blocks were not visible in the conversation timeline. Starting a fresh task on the same Windows machine made the same fenced-code test render correctly. The behavior appears task- or session-dependent.

What steps can reproduce the bug?

1. Open a Codex Desktop task.
2. Emit an inline Markdown item containing a fenced code block.
3. Observe whether the code block is visible in the timeline.
4. Start a fresh task on the same machine and repeat the same test.

Observed: the predecessor task omitted the visible block; the fresh task rendered it.

What is the expected behavior?

Valid inline fenced Markdown should render visibly and consistently in every Codex Desktop task.

Additional information

Impact: technical output can be invisible to the user even when the agent successfully emits valid Markdown, making review and delivery unreliable. No logs, session identifiers, or personal data are included. Please investigate whether rendering state is incorrectly scoped to a task or session, and consider exposing a renderer-health indicator or recovery path.

Start by reproducing the inline fenced-Markdown steps in a predecessor task and a fresh Codex Desktop task on Windows, then compare whether the rendering state differs by task or session. Done means valid fenced code blocks appear consistently in every task; the payload names no files or tests to target.