Cloud review (/code-review ultra) fails consistently on large repo: "all review agents terminated before completing"

Description

/code-review ultra <PR#> (ultrareview) fails identically on back-to-back runs against a large private monorepo. Both cloud sessions end with:

Cloud review failed: orchestrator reported an error: Review failed: all review agents terminated before completing. See session logs for details.

No findings are produced. Two of three free ultrareview runs were consumed by these failures.

Context that suggests a size-related, deterministic failure

• Running local /ultrareview in the same repo is refused with "Repo is too large. Push a PR and use /code-review ultra <PR#> instead." — but the recommended cloud path then dies too, so large repos currently have no working ultra path.
• Repo .git is ~414 MB.
• Target PR: 110 changed files, +5056/−2026.
• Failure reproduced twice with identical error text (2026-08-21 and 2026-08-22), which does not look transient.

Sessions (for internal log lookup)

• session_01VuarTMpNToivmz9Td4mdnZ
• session_01QRDTNfYSH2YyMKomF9xMri

Environment

• Claude Code CLI 2.1.226, macOS (Darwin 25.6.0)
• Launched from terminal in a git worktree of the repo; GitHub connection freshly re-established via /web-setup before the runs

Expected

Cloud review completes (or fails with an actionable message — e.g. "repo/PR too large for cloud review" — before consuming a metered run).

Actual

All review agents terminate before completing; run is consumed; no findings.