[BUG] /ultrareview consumed free run despite 30-minute timeout failure

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Version: 2.1.112 (VS Code extension + CLI, macOS)
Plan: Claude Max 5x
Feature: /ultrareview (research preview)

Bug: Free ultrareview run was consumed despite complete remote session failure with zero findings returned.

Reproduction:

1. cd into a fresh clone of github.com/Tracy625/cq-collab
2. Launch claude CLI
3. Run /ultrareview 44 (PR: +4710/-424, 59 commits)
4. Confirmation dialog showed "Free runs remaining: 3/3"
5. Session launched: session_014tuJwbYUNHRRornrL6vJV5
6. After ~30 minutes, session returned:

"Remote review failed: remote session exceeded 30 minutes"
"Remote review failed (session exceeded 30 minutes). Retry /ultrareview 44, or run /review locally instead."

1. No findings were delivered.
2. Free run count dropped from 3/3 to 2/3.

A second attempt (session_01B4qAupJzBGGhJ86ScFzghA) was launched to verify, consuming another free run. Count is now tracking toward 1/3 if the second session also times out.

What Should Happen?

When a remote ultrareview session fails with a timeout or any system-side error and returns no findings, the failed session should not be counted against the user's free run allotment. Free runs should only be deducted when the review successfully delivers findings to the CLI.

Additionally, the 30-minute hard timeout should be documented in the /ultrareview docs so users can make informed decisions about PR size before launching. The current docs state "10 to 20 minutes" as typical duration with no mention of a hard cap.

Requested resolution:

1. Refund the consumed free run(s) for the failed session(s) on this account.
2. Consider increasing the 30-minute timeout for large PRs, OR surface an up-front PR-size check before the confirmation dialog so users can split the PR or fall back to /review before consuming a free run.

Error Messages/Logs

Steps to Reproduce

1. Ensure Claude Code CLI v2.1.86+ is installed (tested on v2.1.112) and you are on Max 5x plan with 3/3 ultrareview free runs available.

1. Clone a sufficiently large PR target repo:

cd ~/Desktop
git clone https://github.com/Tracy625/cq-collab.git cq-collab-review
cd cq-collab-review

1. Launch Claude Code CLI:

claude

1. In the session, run:

/ultrareview 44
(PR #44 contains +4710/-424 lines across 59 commits — 18 P0/P1 fixes spanning multiple modules)

1. Confirm the launch dialog (shows "Free runs remaining: 3/3", estimated cost 0 credits).

1. Wait approximately 30 minutes. The session returns:

Remote review failed: remote session exceeded 30 minutes
Remote review failed (session exceeded 30 minutes). Retry /ultrareview 44, or run /review locally instead.
No findings are delivered to the CLI.

1. Free run count has dropped from 3/3 to 2/3.

Observations:

• The failure is a system-side timeout, not a user-cancelled or user-error case.
• No partial findings, no diagnostic output — just the timeout message.
• The free run counter decrements identically to a successful run.
• Reproduction is deterministic on large PRs: a second attempt (session_01B4qAupJzBGGhJ86ScFzghA) on the same PR is currently in flight with the same expected outcome.

Relevant session IDs for log lookup:

• session_014tuJwbYUNHRRornrL6vJV5 (first failure)
• session_01B4qAupJzBGGhJ86ScFzghA (second attempt, in progress at time of report)

Note: This bug likely affects any PR large enough to exceed the 30-minute remote timeout. The current /ultrareview docs state typical duration is 10-20 minutes but do not mention the hard cap.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.112

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_