[andreivince]

Issue for this PR

Fixes #50135

Targets v2 per CONTRIBUTING.md. The issue-link warning is the existing workflow problem described in #43964.

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

Mini's session wait can outlive Bun's default fetch idle timeout and terminate with ClientError: Transport. Use the same timeout policy as run for initial and reconnected clients. Explicit cancellation still works, and the prompt is never resubmitted.

How did you verify your code works?

• The subprocess regression passes and fails with unchanged mini code. A delayed local server proves default fetch times out while initial and reconnected mini clients keep waiting; caller cancellation still rejects.
• A consumer probe with the real SDK and session transport completes both delayed turns, renders their responses, and submits each prompt once.
• All 1,362 TUI tests pass, with two skips. Repository lint and all 35 package typechecks pass, with caching disabled.
• Full CLI suite: 284 pass, nine skips, and eight ACP failures. Six failing cases also fail in a PR-free baseline; two do not reproduce in that bounded run. ACP uses a separate client and does not reach the changed mini constructor.

Screenshots / recordings

No visual change.

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

[thdxr]

Thanks @andreivince — this fix is approved. I rebased the branch onto current dev and dropped the two Merge branch 'v2' commits, so the history is now just your commit (authorship kept) plus one small follow-up commit.

The follow-up only touches the test. It no longer requires stderr to be completely empty. Instead it checks that stderr has no Transport: The operation timed out or TimeoutError, and it still checks the exit code and stdout, so unrelated runtime warnings won't make it flaky. I confirmed the test still fails without the timeout: false change. We'll merge once CI is green.

— from 𝕺𝖕𝖊𝖓𝕮𝖔𝖉𝖊