[BUG] /ultrareview consumes free quota even when remote session times out (30min)

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?

Body:
## Summary
A /ultrareview invocation that fails due to "remote session
exceeded 30 minutes"
appears to still count against the 3-free-uses-per-month quota.

## Steps to reproduce

1. Run /ultrareview <base> on a small repo (9 files / ~1300

lines in my case)

1. UI shows: "Free ultrareview 1 of 3" (confirming this was my

first use this period)

1. After ~30+ minutes, task notification arrives:

"Remote review failed: remote session exceeded 30 minutes"

1. No review output is produced
2. Retry the same /ultrareview command → hit a usage limit

message

## Important context

• I have NOT run /ultrareview in any other session this period.

The first attempt today was genuinely my first use, as the "1
of 3" display confirmed.

• Therefore at most 1 free slot should have been consumed (and

arguably zero, since
it was a server-side timeout, not a user action).

## Expected behavior
Failed runs (especially server-side timeouts that the user has no
control over)
should NOT consume the free quota. Or, the CLI should warn the
user before
re-trying that the quota may be consumed on failure.

## Actual behavior

• No review output delivered
• Quota appears consumed (no refund / no retry credit)
• No proactive warning before retry that quota would be consumed

on timeout

• Note: the limit message I hit on retry mentioned "/upgrade",

which suggests
this may be the general plan usage limit rather than the
ultrareview-specific
3-free-uses quota. Please clarify which limit was triggered,
since this affects
whether the timeout itself is the root cause.

## Environment

• Claude Code on Windows 11
• Plan: Pro
• Dates: 2026-04-28 (first attempt timeout) / 2026-04-29 (retry

hit limit) JST

## Suggested fix

• Refund quota on remote-side timeouts (failure code !=

user-cancel)

• Or surface a confirmation prompt: "Previous run timed out —

retrying will
consume another free slot. Continue?"

• Distinguish clearly in the failure message whether the

ultrareview-specific
quota OR the general plan quota was exhausted

What Should Happen?

Failed runs (especially server-side timeouts that the user has no
control over)
should NOT consume the free quota. Or, the CLI should warn the
user before
re-trying that the quota may be consumed on failure.

Error Messages/Logs

Steps to Reproduce

1. Run /ultrareview <base> on a small repo (9 files / ~1300

lines in my case)

1. UI shows: "Free ultrareview 1 of 3" (confirming this was my

first use this period)

1. After ~30+ minutes, task notification arrives:

"Remote review failed: remote session exceeded 30 minutes"

1. No review output is produced
2. Retry the same /ultrareview command → hit a usage limit

message

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

Claude Code Version: 2.1.122

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

③ Additional context

• I have NOT run /ultrareview in any other session this period.

The first attempt today was genuinely my first use, as the "1
of 3" display confirmed.

• Therefore at most 1 free slot should have been consumed (and

arguably zero,
since it was a server-side timeout, not a user action).

• Dates: 2026-04-28 (first attempt timeout) / 2026-04-29 (retry

hit limit) JST

Suggested fix:

• Refund quota on remote-side timeouts (failure code !=

user-cancel)

• Or surface a confirmation prompt before retry: "Previous run

timed out —
retrying will consume another free slot. Continue?"

• Distinguish clearly in the failure message whether the

ultrareview-specific
quota OR the general plan quota was exhausted