/ultrareview crashed before producing findings — request credit refund

Summary

Ran /ultrareview from Claude Code (CLI). The task notification came back as status: completed almost immediately, but the result payload was a JSON error string with no findings:

{"error":"Review crashed before producing findings. See session logs for details."}

Repro

• Command: /ultrareview (no args — branch-mode, current branch vs main)
• Diff scope at launch: ~120 files changed, ~16,084 insertions, ~2,055 deletions
• Free-tier counter at launch: "Free ultrareview 1 of 3"
• Session URL: https://claude.ai/code/session_01RGhBK7iX1au7K1hgRS5Yqe

Expected

Either real review findings, or a graceful "couldn't review (reason), please retry — your free run was not consumed" message.

Actual

The free-tier counter appears to have been decremented even though zero findings were produced. Requesting that the run be credited back since the review never delivered any output.

Notes

The error message points to "session logs" but those aren't surfaced to the client. One plausible cause given the diff size is an implicit per-run size/token cap during the multi-agent fan-out, but the cloud-side stack trace is needed to confirm. Routing this to whoever owns the ultrareview backend is probably the fastest path.