A5 (drift indicator on the Scorecard Dimensions tab) answers a per-suite question: is this judge criterion drifted right now? Useful when you're already debugging a specific suite. Useless when the question is "across every suite I own, which judges should I recalibrate today?"

The Calibration Center on Reviews → Calibration is the cross-suite answer. It reads the same calibration_baselines table A5 reads, but slices across every suite, surfaces only the rows whose baseline is older than the freshness threshold (3 days), and sorts so the most urgent recalibration target sits at the top.

What it does today

• Lists every (suite, criterion) baseline in the workspace from GET /scoring/calibration-baselines.
• Derives days_since_baseline client-side via toOverviewRow so the freshness threshold stays in sync with the Smart suggestions strip (RULE_DRIFT_STALE_BASELINE, also 3 days).
• Renders the drifted list with per-criterion stats: Pearson r, MAE, Cohen's κ, sample count.
• Each row has an Open in Scorecard action that lands on the suite's Experiment Health tab — the same A5 destination the Smart suggestions strip uses.
• A summary line reports how many baselines are still inside the freshness window (and therefore not surfaced as drifted), so the operator sees the full denominator.

What is honestly deferred

The §4.2 mockup names two further row actions (retire and fold into rubric) that require backend signal Winnow doesn't have yet: a retired_at column on baselines, and a rubric-fold ingest route on EvalBuilder criteria. Both ship as named follow-ups in docs/UNBUILT-FEATURES.md; neither is rendered as a placeholder button in the UI.

Bulk actions are visible-but-disabled with a tooltip pointing at the unbuilt endpoint, because that's a better signal than hiding the buttons (an operator working through 20 drifted judges knows the workflow exists and that we know they need it).

"Days drifted" is currently days_since_baseline (a staleness proxy). A true first_drifted_at requires a typed drift-state column on calibration_baselines. The page calls this out in a wiki link so an operator doesn't mistake the proxy for a measured signal.

Why "cross-suite triage" deserves a tab

A new tab on Reviews is the kind of UI we'd normally rule out — but the cross-suite roll-up has no existing home:

• Per-suite drift lives on the Scorecard → Dimensions tab (A5). That's a single-suite diagnostic.
• Per-row HIL labelling lives on Reviews → Human input (A4). That's per-request.
• Per-suite scaffold history lives on /scaffolds (B5). That's per-suite again.

None of those tabs answer the multi-suite, multi-criterion, "what should I do next" question. The Calibration Center does, in a single triage view, and routes the operator to the existing per-suite recalibration affordance with one click.