@@ -0,0 +1,250 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+/**
+ * [#25467 cut 2] The metadata diagnostics sweep of
+ * `ObjectStackProtocolImplementation` — `getMetaDiagnostics`, which sweeps the
+ * metadata types a caller asks about, reports the entries that fail spec
+ * validation and the per-type stats beside them (`GET /api/v1/meta/diagnostics`)
+ * — moved out of `protocol.ts` whole.
+ *
+ * A MOVE, not a redesign. The function below is the body of the
+ * `ObjectStackProtocolImplementation` method of the same name as it stood,
+ * re-indented one level and otherwise unedited: the same reads, the same
+ * findings, the same counts.
+ *
+ * ## How the moved body reaches the class
+ *
+ * The shape #23050 set for `rest-server.ts` (its `data-action-routes.ts`
+ * header carries the full reasoning), as `protocol-batch-data.ts` applies it.
+ * The class keeps the method — the published `dist/*.d.ts` lists its name,
+ * docblock, signature and position — and its body is one literal of exactly
+ * the {@link MetaDiagnosticsHost} members it reads, plus a `.call`. The method
+ * is no longer `async`: it returns the moved function's own promise, so a
+ * caller awaits the same promise the body settles, with no extra wrapper. Each
+ * member is handed over as a closure that calls it on the protocol at call
+ * time, so a test double or spy set on the instance is still the one the body
+ * reaches.
+ *
+ * The one module-level declaration the body reads, `ShippingPackages`, stays in
+ * `protocol.ts`, where code that did not move reads it too, and is imported
+ * from there as a type. `index.ts` names what it publishes from `protocol.ts`
+ * one by one, so the `export` it gained for this keeps it off the published
+ * surface, and nothing re-exports this module, so the published surface does
+ * not move.
+ */
+
+import type { ObjectStackProtocolImplementation, ShippingPackages } from './protocol.js';
+import { resolveItemLock } from './item-lock.js';
+import { DEFAULT_METADATA_TYPE_REGISTRY, getMetadataTypeSchema } from '@objectstack/spec/kernel';
+import { computeMetadataDiagnostics } from './metadata-diagnostics.js';
+import type { MetadataDiagnostics } from './metadata-diagnostics.js';
+
+/**
+ * The `ObjectStackProtocolImplementation` members the function below reads —
+ * every one of them, and no other; its `this` is the `Pick` of them. Each is
+ * typed AS the class member (an indexed access on the class), so a change to a
+ * member's signature on the class is a change here too, and naming a member
+ * the class does not have is a compile error.
+ */
+export interface MetaDiagnosticsHost {
+    artifactLockLayerAt: ObjectStackProtocolImplementation['artifactLockLayerAt'];
+    getMetaItems: ObjectStackProtocolImplementation['getMetaItems'];
+    isArtifactBacked: ObjectStackProtocolImplementation['isArtifactBacked'];
+    servedLockState: ObjectStackProtocolImplementation['servedLockState'];
+    shippingPackagesOf: ObjectStackProtocolImplementation['shippingPackagesOf'];
+}
+
+/** The body of `ObjectStackProtocolImplementation#getMetaDiagnostics`; the method's docblock there is its contract. */
+export async function getMetaDiagnostics(
+    this: Pick<MetaDiagnosticsHost, 'artifactLockLayerAt' | 'getMetaItems' | 'isArtifactBacked' | 'servedLockState' | 'shippingPackagesOf'>,
+    request: {
+    type?: string;
+    severity?: 'error' | 'warning';
+    packageId?: string;
+} = {}): Promise<{
+    entries: Array<{ type: string; name: string; diagnostics: MetadataDiagnostics }>;
+    total: number;
+    scannedTypes: number;
+    scannedItems: number;
+    /**
+     * Per-type aggregate stats — count of items and the list of
+     * packages contributing to each type. Computed in the same
+     * sweep so the Studio directory page can render tile counts
+     * and a package filter in one round-trip. `locked` is the number
+     * of items whose read envelope reports `lock` other than `'none'`
+     * — the same derivation `getMetaItem` publishes (#21694).
+     */
+    stats: Record<string, { count: number; locked: number; packages: string[] }>;
+}> {
+    const includeWarnings = request.severity === 'warning';
+    const targetTypes = request.type
+        ? [request.type]
+        : DEFAULT_METADATA_TYPE_REGISTRY
+            .filter((e) => getMetadataTypeSchema(e.type))
+            .map((e) => e.type);
+
+    const entries: Array<{ type: string; name: string; diagnostics: MetadataDiagnostics }> = [];
+    const stats: Record<string, { count: number; locked: number; packages: string[] }> = {};
+    let scannedItems = 0;
+
+    for (const t of targetTypes) {
+        let listed: any;
+        try {
+            listed = await this.getMetaItems({
+                type: t,
+                packageId: request.packageId,
+            } as any);
+        } catch (error) {
+            // [#8855] TWO different failures used to leave through this
+            // `continue`, and only one of them is benign.
+            //
+            // BENIGN, and the reason the comment named — it is real, which
+            // is why this `catch` is narrowed rather than deleted: the type
+            // is not listable in THIS kernel scope. Skipping it is the
+            // truth, and a scope that cannot enumerate one type must not
+            // fail the whole governance sweep.
+            //
+            // NOT BENIGN: the read already ran the platform's single
+            // discrimination — `getMetaItems` classifies by error TYPE
+            // through {@link rethrowUnlessMetadataStoreUnprovisioned},
+            // returns normally for the one benign driver reason
+            // (`isMissingTableError`: `sys_metadata` not provisioned yet,
+            // so `items: []` IS the truth), and throws
+            // {@link metadataStoreUnavailableError} — a 503 — for every
+            // other read failure. That 503 says "whether rows exist is
+            // UNKNOWN". Swallowing it here put the emptiness straight back
+            // one layer up, and then published it as a NUMBER:
+            //
+            //   * `stats[t]` was never written, so the type is ABSENT from
+            //     the response — not zero, absent — byte-shaped like an
+            //     environment that declares none of it, and the Studio
+            //     directory tile the field exists to feed simply loses it;
+            //   * `total` counts entries that FAILED validation, and a
+            //     store nobody can read contributes none — so a diagnostics
+            //     endpoint whose whole job is reporting problems answered
+            //     `0 problems` precisely when it could not read anything.
+            //
+            // ADR-0110 D3 is the rule: a miss and an outage are different
+            // facts with opposite dispositions ("nothing to fix" vs "the
+            // backend is down, retry"), and a consumer must never read one
+            // as the other. #5108 restored it in `DatabaseLoader`, #5532 in
+            // `getMetaItems` itself — this is the same rule one layer up,
+            // on the consumer that turns the read into a published count.
+            //
+            // Classification is by the envelope the PRODUCER already built,
+            // not by a second reading of the driver error. Re-running
+            // `isMissingTableError` / `rethrowUnlessMetadataStoreUnprovisioned`
+            // on what arrives here would re-wrap an already-shaped 503 in
+            // another one and displace the driver error riding as `cause` —
+            // the object `logWithheldServerFault` prints for the operator
+            // (#5437). The original is rethrown UNCHANGED, so the envelope
+            // that reaches the REST boundary is the one #5532 shaped.
+            //
+            // The test is `status === 503` rather than the narrower
+            // `code === 'SERVICE_UNAVAILABLE'` on purpose, and in the
+            // direction `rethrowUnlessMetadataStoreUnprovisioned`'s own doc
+            // argues: a false "benign" silently mis-answers a question
+            // nobody can re-ask, while a false "real" costs one 503 the
+            // caller can retry.
+            if ((error as { status?: unknown } | null | undefined)?.status === 503) throw error;
+            // [#8924] TWO error classes leave through this catch — the 503
+            // above and the 400 below — under ONE rule: an error the
+            // producer already stamped with a status is a CLASSIFIED
+            // VERDICT, and this sweep may not overrule a verdict into
+            // silence. Everything else is a genuine listing failure nobody
+            // classified, and for those the benign skip below is the ruled
+            // disposition (maintainer, 2026-08-20, option 1).
+            //
+            //  * 503 (#8855, above): the store outage — "whether rows
+            //    exist is UNKNOWN". Swallowing it published an outage as
+            //    "0 problems".
+            //  * 400 (#8924, here): the request-classification refusal
+            //    `canonicalizeMetaRequestType` raises INSIDE `getMetaItems`
+            //    (`status: 400`, `code: 'INVALID_REQUEST'`) for an
+            //    unrecognised spelling of a DECLARED type (`fieldes`,
+            //    `viewes`, `capabilitys`, …). Only the `request.type` arm
+            //    can produce it — the full-sweep arm's `targetTypes` come
+            //    canonical out of the registry — so rethrowing it cannot
+            //    fail a whole-corpus sweep. Swallowing it answered
+            //    `200 {"scannedTypes":1,"stats":{}}` — "scanned 1 type,
+            //    no issues" — to a spelling every sibling `/meta` door
+            //    refuses with a 400 that names both accepted spellings
+            //    (measured: 27 such spellings on a booted kernel, while
+            //    objectui's DiagnosticsPage paints that payload as "All
+            //    clear"). The producer classified the caller's mistake;
+            //    the caller gets to hear it. Declared = enforced.
+            //
+            // What still takes the `continue`: a listing failure with no
+            // status, or any status that is not one of the two verdicts
+            // above (the registry listing is not a guarded read, so its
+            // failures arrive unclassified). That skip is genuine — a
+            // kernel scope that cannot enumerate one type must not fail
+            // the whole governance sweep — and measured to have NO
+            // producer in kernel scope today (six booted scopes, zero
+            // skips), which is exactly why it stays a guard rather than
+            // becoming a payload field (`skippedTypes` was considered and
+            // ruled out as motiveless).
+            //
+            // By `status` and not by `code`, mirroring the 503 test above
+            // and its direction argument: a false "real" is loud and
+            // correctable at the caller; a false "benign" silently
+            // mis-answers a question nobody re-asks.
+            if ((error as { status?: unknown } | null | undefined)?.status === 400) throw error;
+            continue;
+        }
+        const items: any[] = Array.isArray(listed?.items)
+            ? listed.items
+            : Array.isArray(listed)
+                ? listed
+                : [];
+        const pkgSet = new Set<string>();
+        let lockedCount = 0;
+        let shipping: ShippingPackages | undefined;
+        for (const item of items) {
+            scannedItems += 1;
+            const pkg = (item?._packageId ?? null) as string | null;
+            if (pkg) pkgSet.add(pkg);
+            // [#21694] Counted from the envelope's own derivation
+            // ({@link servedLockState}), never from the declared `_lock`
+            // alone: a packaged base the write doors refuse reads locked on
+            // its item envelope, so the per-type tile counts it too. The
+            // derivation reads the registry only — no store read per item.
+            // [#21738] Its lock is the one item-lock resolution
+            // ({@link resolveItemLock}) over the item's artifact layer —
+            // [#21803] every installed package that ships the name, from
+            // the address the list's own merge used (ADR-0048 package
+            // scope) — and the document the list serves for it.
+            const itemName = typeof item?.name === 'string' ? item.name : '';
+            shipping ??= this.shippingPackagesOf(t);
+            const itemLock = resolveItemLock({
+                artifact: this.artifactLockLayerAt({
+                    type: t,
+                    name: itemName,
+                    packageId: request.packageId ?? (item?._packageId as string | undefined),
+                }, shipping),
+                overlay: item,
+            });
+            const served = this.servedLockState(t, itemName, item, this.isArtifactBacked(t, itemName), itemLock);
+            if (served.lock !== 'none') lockedCount += 1;
+            const diag: MetadataDiagnostics | undefined =
+                item?._diagnostics ?? computeMetadataDiagnostics(t, item);
+            if (!diag) continue;
+            if (diag.valid && !includeWarnings) continue;
+            if (diag.valid && includeWarnings && !diag.warnings?.length) continue;
+            entries.push({
+                type: t,
+                name: typeof item?.name === 'string' ? item.name : '<unknown>',
+                diagnostics: diag,
+            });
+        }
+        stats[t] = { count: items.length, locked: lockedCount, packages: [...pkgSet].sort() };
+    }
+
+    return {
+        entries,
+        total: entries.length,
+        scannedTypes: targetTypes.length,
+        scannedItems,
+        stats,
+    };
+}