Clara nightly behavior evals — honest green

The scheduled Clara Behavior Evals nightly on main has failed every night Oct 15-24: seven domains red every night plus a rotating fringe. Check common cause before per-domain fixes. Done is an observed scheduled nightly concluding success, or every red domain carrying a measured floor plus a tracked-gap row.

Status is DERIVED from each row’s own PR, decision and prod-evidence links by bin/refresh-tracker clara-nightly-evals-green — never typed. Last refreshed 2026-10-31 01:33Z.

Every row opens: when it was added, why it exists, and the links that prove it. Rows waiting on a human carry a Decide → link straight to the answerable question on the decisions page. Statuses are re-derived from live PR, decision and prod-evidence state by bin/refresh-trackers; nothing here is typed by hand. Merged and shipped both count as done (Gera, 2026-10-19); a prod_evidence permalink is what moves a row from Merged to Shipped.