Fix operator-state null capabilityNames crash and stopped-container gating

Issue #1621: saving only "intentionally offline" left the SQLite
auto_remediation_policy_json column NULL, and the read path only
normalized the policy inside the non-NULL branch, so the API served
"capabilityNames": null. The section's dirty-state memo spread that
value during render and crashed the expanded row into the route error
boundary. Normalize the policy unconditionally after the scan so the
wire always carries [], type capabilityNames as nullable in the TS API
surface, and guard the spread in the component.

Issue #1622: the inline table-row presentation gated the entire
operator-overrides section on the resource exposing an auto-authorizable
capability, so stopped Docker containers (which only expose start,
never auto-authorized) lost intentionally-offline and
never-auto-remediate exactly when they matter most. Render the section
whenever the resource has an id; the automatic-actions block already
self-gates on eligible capabilities. Also add min-w-0 to the section's
flex text columns and overflow-x-clip to the inline row content box so
the long never-auto-remediate copy wraps instead of escaping the row
border below the lg breakpoint.

Regression coverage: issue-named Go test for the NULL-column read path
and a render test feeding the section the real pre-fix wire payload
with capabilityNames: null.

Fixes #1621
Fixes #1622

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

docs/release-control/v6/internal/subsystems/api-contracts.md

@@ -4458,6 +4458,12 @@ normalizes the server's `404 operator_state_not_set` response into
 than a thrown error; non-404 errors propagate. The PUT path
 percent-encodes the canonical resource id segment so colon-bearing
 ids round-trip safely through URL routing.
+Operator-state reads must keep `autoRemediationPolicy.capabilityNames`
+iterable. SQLite rows whose policy column is `NULL` normalize to an empty
+capability list before JSON serialization, while the TS compatibility type and
+rendering path continue to tolerate `null` from pre-fix or cached payloads.
+Saving only `intentionallyOffline`, `neverAutoRemediate`, priority, or note
+must therefore never produce a drawer-breaking operator-state response.
 
 The router wires the operator-state adapter into the findings runtime
 at startup: `internal/api/router.go` calls

docs/release-control/v6/internal/subsystems/frontend-primitives.md

@@ -3688,7 +3688,10 @@ Inline detail table rows are also registry-backed. `InlineDetailTableRow`
 owns the row/cell/content shell and row-click containment for platform,
 workload, and infrastructure inline drawers; callers may pass row-specific
 `data-*` attributes, colspan, and content classes, but they must not recreate
-the surface-alt detail row shell locally.
+the surface-alt detail row shell locally. The content shell must clip
+horizontal paint below the large breakpoint without becoming a scroll
+container, then restore visible overflow for the static desktop layout, so
+long operator-state copy cannot escape the shared row border.
 Inline detail section content is registry-backed separately from the row shell.
 `DetailSectionTable`, `InlineDetailPanel`, and `detailSectionModel.ts` own
 detail row compaction, section-table rendering, value-tone classes, and the

docs/release-control/v6/internal/subsystems/unified-resources.md

@@ -1790,7 +1790,11 @@ capability is never eligible, low-risk eligible, or elevated eligible; it does
 not lower `MinimumApprovalLevel`. The first eligible vertical is Docker/Podman
 container `restart` at `low_risk`; unspecified capabilities normalize to never.
 `NormalizeResourceOperatorState` trims whitespace, de-duplicates capability
-names, and lower-cases the criticality value before persistence. The
+names, guarantees a non-nil empty capability list, and lower-cases the
+criticality value before persistence. SQLite reads must apply that
+normalization even when `auto_remediation_policy_json` is `NULL`, so an
+operator state containing only intentionally-offline or another independent
+override serializes `capabilityNames: []` rather than `null`. The
 `ResourceStore` interface now exposes
 `GetResourceOperatorState`, `SetResourceOperatorState`, and
 `ClearResourceOperatorState`; both the SQLite (table
@@ -1821,7 +1825,10 @@ overview tab, which routes through the canonical TS client at
 `frontend-modern/src/api/resourceOperatorState.ts` to
 `/api/resources/{id}/operator-state`. Alongside the existing operator-state
 controls, the drawer offers automatic actions only for backend-advertised
-eligible capabilities. Enabling it requires selecting exact capabilities and
+eligible capabilities. The operator-state section itself remains available for
+every canonical resource id, including stopped containers and resources with
+no auto-authorizable capability; only the automatic-actions subsection is
+capability-gated. Enabling it requires selecting exact capabilities and
 may add a recurring daily time/timezone window; copy must state that this is a
 resource allowlist and that the tenant Patrol mode remains the upper bound.
 The UI must never infer eligibility from capability names, severity, or the

frontend-modern/src/api/__tests__/resourceOperatorState.test.ts

@@ -44,6 +44,23 @@ describe('resourceOperatorState api', () => {
     await expect(getResourceOperatorState('vm:101')).resolves.toBeNull();
   });
 
+  it('preserves a legacy null capability list for compatibility consumers to normalize', async () => {
+    apiFetchJSONMock.mockResolvedValueOnce({
+      canonicalId: 'docker:host-1/ct-nginx',
+      intentionallyOffline: true,
+      neverAutoRemediate: false,
+      autoRemediationPolicy: {
+        enabled: false,
+        capabilityNames: null,
+      },
+      setAt: '2026-07-20T10:00:00Z',
+    } satisfies ResourceOperatorState);
+
+    const result = await getResourceOperatorState('docker:host-1/ct-nginx');
+
+    expect(result?.autoRemediationPolicy?.capabilityNames).toBeNull();
+  });
+
   it('rethrows non-404 errors so the caller can surface them to the operator', async () => {
     apiFetchJSONMock.mockRejectedValueOnce(
       Object.assign(new Error('Internal server error'), { status: 500 }),

frontend-modern/src/api/resourceOperatorState.ts

@@ -17,7 +17,12 @@ export interface AutoRemediationWindow {
 
 export interface AutoRemediationPolicy {
   enabled: boolean;
-  capabilityNames: string[];
+  /**
+   * Nullable on the wire: rows persisted before the store normalized the
+   * policy on read carry `"capabilityNames": null` (issue #1621). Treat
+   * null the same as [].
+   */
+  capabilityNames: string[] | null;
   window?: AutoRemediationWindow;
 }
 

frontend-modern/src/components/Infrastructure/ResourceDetailDrawerOverviewTab.tsx

@@ -355,11 +355,6 @@ export const ResourceDetailDrawerOverviewTab: Component<ResourceDetailDrawerOver
     drawer.resourceTimelineCount() > 0 ||
     Boolean(drawer.facetBundleError());
   const shouldRenderOperationalGovernanceSections = () => !compactTableRow();
-  const shouldRenderOperatorStateSection = () =>
-    !compactTableRow() ||
-    resource.capabilities?.some(
-      (capability) => capability.autoAuthorization && capability.autoAuthorization !== 'never',
-    );
   const shouldRenderActionHistorySection = () =>
     drawer.actionAuditAvailable() &&
     (!compactTableRow() ||
@@ -697,8 +692,13 @@ export const ResourceDetailDrawerOverviewTab: Component<ResourceDetailDrawerOver
             never auto-remediate) sits next to the action history so
             the "what overrides has the operator set" and "what actions
             has Pulse taken" stories read together. The section is
-            self-fetching keyed on the canonical resource id. */}
-        <Show when={shouldRenderOperatorStateSection() && resource.id}>
+            self-fetching keyed on the canonical resource id. Rendered
+            for every resource with an id — the operator overrides
+            (intentionally offline, never auto-remediate, note, priority)
+            don't depend on capabilities; only the automatic-actions
+            block does, and it gates itself inside the section
+            (issue #1622). */}
+        <Show when={resource.id}>
           <ResourceOperatorStateSection
             resourceId={resource.id}
             capabilities={resource.capabilities}

frontend-modern/src/components/Infrastructure/ResourceOperatorStateSection.tsx

@@ -139,7 +139,9 @@ export const ResourceOperatorStateSection: Component<ResourceOperatorStateSectio
       capabilityNames: [],
     };
     const selected = [...autoCapabilities()].sort();
-    const persistedSelected = [...persistedAuto.capabilityNames].sort();
+    // capabilityNames can arrive as null from rows persisted before the
+    // policy column was always written (issue #1621) — never assume [].
+    const persistedSelected = [...(persistedAuto.capabilityNames ?? [])].sort();
     const windowChanged = autoWindowEnabled()
       ? !persistedAuto.window ||
         persistedAuto.window.timezone !== autoWindowTimezone() ||
@@ -536,7 +538,7 @@ export const ResourceOperatorStateSection: Component<ResourceOperatorStateSectio
           = +1h) or with the persisted window when one exists. */}
       <Show when={!schedulerOpen()}>
         <div class="flex items-center justify-between gap-3 pt-2 border-t border-border-subtle">
-          <div class="flex-1">
+          <div class="min-w-0 flex-1">
             <label class="text-sm font-medium text-base-content">Maintenance window</label>
             <p class="text-[11px] text-muted mt-0.5 leading-tight">
               Suspend findings on this resource for a defined window. Useful for scheduled upgrades,
@@ -673,7 +675,7 @@ export const ResourceOperatorStateSection: Component<ResourceOperatorStateSectio
       <Show when={eligibleAutoCapabilities().length > 0}>
         <div class="space-y-3 border-t border-border-subtle pt-3" aria-label="Automatic actions">
           <div class="flex items-start justify-between gap-3">
-            <div class="flex-1">
+            <div class="min-w-0 flex-1">
               <label class="text-sm font-medium text-base-content">Automatic actions</label>
               <p class="mt-0.5 text-[11px] leading-tight text-muted">
                 Choose exactly which actions Patrol may run on this resource. Your Patrol mode, live
@@ -780,7 +782,7 @@ export const ResourceOperatorStateSection: Component<ResourceOperatorStateSectio
       </Show>
 
       <div class="flex items-start justify-between gap-3">
-        <div class="flex-1">
+        <div class="min-w-0 flex-1">
           <label class="text-sm font-medium text-base-content">Intentionally offline</label>
           <p class="text-[11px] text-muted mt-0.5 leading-tight">
             Suppress findings on this resource. Use when a workload is deprecated, a dev environment
@@ -795,7 +797,7 @@ export const ResourceOperatorStateSection: Component<ResourceOperatorStateSectio
       </div>
 
       <div class="flex items-start justify-between gap-3 pt-2 border-t border-border-subtle">
-        <div class="flex-1">
+        <div class="min-w-0 flex-1">
           <label class="text-sm font-medium text-red-700 dark:text-red-400">
             Never auto-remediate
           </label>

frontend-modern/src/components/Infrastructure/__tests__/ResourceDetailDrawer.history.test.tsx

@@ -366,7 +366,10 @@ describe('ResourceDetailDrawer change history section', () => {
     expect(screen.getByText('Current state')).toBeInTheDocument();
     expect(screen.getByText('Identity')).toBeInTheDocument();
     expect(screen.queryByText('Change history')).not.toBeInTheDocument();
-    expect(screen.queryByText('Operator overrides')).not.toBeInTheDocument();
+    // Operator overrides render in every presentation (issue #1622) —
+    // intentionally-offline and never-auto-remediate must stay reachable
+    // even for resources with no auto-authorizable capabilities.
+    expect(screen.getByText('Operator overrides')).toBeInTheDocument();
     expect(screen.queryByText('Maintenance verification')).not.toBeInTheDocument();
     expect(screen.queryByText('Action history')).not.toBeInTheDocument();
     expect(screen.queryByText('Refreshing changes...')).not.toBeInTheDocument();

frontend-modern/src/components/Infrastructure/__tests__/ResourceOperatorStateSection.render.test.tsx

@@ -0,0 +1,94 @@
+import { describe, expect, it, vi } from 'vitest';
+import { render, screen, waitFor } from '@solidjs/testing-library';
+
+import type { ResourceCapability } from '@/types/resource';
+
+const apiClientMock = vi.hoisted(() => ({
+  apiFetchJSON: vi.fn(),
+}));
+
+vi.mock('@/utils/apiClient', async (importOriginal) => {
+  const actual = await importOriginal<typeof import('@/utils/apiClient')>();
+  return {
+    ...actual,
+    apiFetchJSON: apiClientMock.apiFetchJSON,
+  };
+});
+
+import { ResourceOperatorStateSection } from '@/components/Infrastructure/ResourceOperatorStateSection';
+
+// Regression for issue #1621: rows saved with only "intentionally offline"
+// left the auto_remediation_policy_json column NULL, and the pre-fix read
+// path skipped normalization for NULL columns, so the wire carried
+// "capabilityNames": null. The section's dirty-state memo spread that value
+// during render, throwing "null is not iterable" straight into the route
+// error boundary. These tests feed the section the real pre-fix wire
+// payload through the canonical API client to pin the null tolerance.
+const OPERATOR_STATE_WITH_NULL_CAPABILITIES = JSON.parse(
+  `{
+    "canonicalId": "docker:host-1/ct-nginx",
+    "intentionallyOffline": true,
+    "neverAutoRemediate": false,
+    "autoRemediationPolicy": {"enabled": false, "capabilityNames": null},
+    "setAt": "2026-07-20T10:00:00Z",
+    "setBy": "admin"
+  }`,
+) as unknown;
+
+const RESTART_CAPABILITY: ResourceCapability[] = [
+  {
+    name: 'restart',
+    type: 'docker',
+    description: 'Restart the container',
+    autoAuthorization: 'low_risk',
+  },
+];
+
+describe('ResourceOperatorStateSection render with capabilityNames: null', () => {
+  it('renders the persisted record without crashing and reports a clean (not dirty) state', async () => {
+    apiClientMock.apiFetchJSON.mockResolvedValue(OPERATOR_STATE_WITH_NULL_CAPABILITIES);
+
+    render(() => (
+      <ResourceOperatorStateSection
+        resourceId="docker:host-1/ct-nginx"
+        capabilities={RESTART_CAPABILITY}
+      />
+    ));
+
+    // The section itself must be up (pre-fix, the throw escaped render
+    // and nothing below the header survived).
+    expect(screen.getByText('Operator overrides')).toBeTruthy();
+    expect(screen.getByText('Intentionally offline')).toBeTruthy();
+    expect(screen.getByText('Never auto-remediate')).toBeTruthy();
+
+    // "Clear all overrides" only renders for persisted() && !isDirty() —
+    // reaching it proves the dirty-state memo evaluated the null
+    // capabilityNames without throwing and treated it as [].
+    await waitFor(() => {
+      expect(screen.getByText('Clear all overrides')).toBeTruthy();
+    });
+    expect(screen.getByText(/Set by admin/)).toBeTruthy();
+    expect(screen.queryByText('Save overrides')).toBeNull();
+  });
+
+  it('hydrates a null capability allowlist as an empty selection in the automatic-actions block', async () => {
+    apiClientMock.apiFetchJSON.mockResolvedValue(OPERATOR_STATE_WITH_NULL_CAPABILITIES);
+
+    render(() => (
+      <ResourceOperatorStateSection
+        resourceId="docker:host-1/ct-nginx-2"
+        capabilities={RESTART_CAPABILITY}
+      />
+    ));
+
+    await waitFor(() => {
+      expect(screen.getByText('Clear all overrides')).toBeTruthy();
+    });
+
+    // The automatic-actions block renders (the resource has an eligible
+    // capability) with its toggle off and no phantom selection derived
+    // from the null allowlist.
+    expect(screen.getByText('Automatic actions')).toBeTruthy();
+    expect(screen.queryByText('Allowed actions')).toBeNull();
+  });
+});

frontend-modern/src/components/Infrastructure/__tests__/ResourceOperatorStateSection.test.ts

@@ -197,12 +197,16 @@ describe('ResourceDetailDrawerOverviewTab integration', () => {
     expect(operatorIndex).toBeLessThan(historyIndex);
   });
 
-  it('keeps eligible automatic actions reachable from compact platform rows', () => {
-    expect(overviewTabSource).toContain('shouldRenderOperatorStateSection');
-    expect(overviewTabSource).toContain('capability.autoAuthorization');
-    expect(overviewTabSource).toContain("capability.autoAuthorization !== 'never'");
-    expect(overviewTabSource).toContain(
-      '<Show when={shouldRenderOperatorStateSection() && resource.id}>',
-    );
+  it('renders operator overrides for every resource regardless of capability eligibility', () => {
+    // Issue #1622: the section used to be gated on the resource exposing
+    // an auto-authorizable capability, which hid intentionally-offline /
+    // never-auto-remediate exactly when they matter most (a stopped
+    // container only exposes `start`, which is never auto-authorized).
+    // The overrides don't depend on capabilities — only the
+    // automatic-actions block does, and it self-gates inside the section
+    // via eligibleAutoCapabilities.
+    expect(overviewTabSource).toContain('<Show when={resource.id}>');
+    expect(overviewTabSource).not.toContain('shouldRenderOperatorStateSection');
+    expect(sectionSource).toContain('<Show when={eligibleAutoCapabilities().length > 0}>');
   });
 });

frontend-modern/src/components/shared/InlineDetailTableRow.tsx

@@ -3,8 +3,11 @@ import { splitProps, type JSX } from 'solid-js';
 import { TableCell, TableRow, type TableRowProps } from './Table';
 
 export const INLINE_DETAIL_TABLE_CELL_CLASS = 'p-0 border-b border-border bg-surface-alt';
+// overflow-x-clip keeps over-wide content from painting outside the row's
+// border below the lg breakpoint (issue #1622) without creating a scroll
+// container the way overflow-hidden would.
 export const INLINE_DETAIL_TABLE_CONTENT_CLASS =
-  'sticky left-0 max-w-[calc(100vw-3.5rem)] px-2 py-3 sm:px-4 sm:py-4 lg:static lg:max-w-none';
+  'sticky left-0 max-w-[calc(100vw-3.5rem)] overflow-x-clip px-2 py-3 sm:px-4 sm:py-4 lg:static lg:max-w-none lg:overflow-x-visible';
 
 export interface InlineDetailTableRowProps extends TableRowProps {
   cellId?: string;

frontend-modern/src/components/shared/SharedPrimitives.guardrails.test.ts

@@ -7408,6 +7408,8 @@ describe('shared primitive guardrails', () => {
     expect(registeredRule?.canonical?.export).toBe('InlineDetailTableRow');
     expect(inlineDetailTableRowSource).toContain('INLINE_DETAIL_TABLE_CELL_CLASS');
     expect(inlineDetailTableRowSource).toContain('INLINE_DETAIL_TABLE_CONTENT_CLASS');
+    expect(inlineDetailTableRowSource).toContain('overflow-x-clip');
+    expect(inlineDetailTableRowSource).toContain('lg:overflow-x-visible');
     expect(inlineDetailTableRowSource).toContain('event.stopPropagation()');
     expect(registeredRule?.requiredConsumers?.map((consumer) => consumer.path)).toEqual(
       expect.arrayContaining([

internal/unifiedresources/resource_operator_state_policy_test.go

@@ -1,6 +1,8 @@
 package unifiedresources
 
 import (
+	"encoding/json"
+	"strings"
 	"testing"
 	"time"
 )
@@ -84,3 +86,41 @@ func TestAutoRemediationPolicySQLiteRoundTrip(t *testing.T) {
 		t.Fatalf("round-trip policy = %#v", got.AutoRemediationPolicy)
 	}
 }
+
+func TestSQLiteOperatorStateNullPolicyColumnNormalizesOnRead(t *testing.T) {
+	store, err := NewSQLiteResourceStore(t.TempDir(), defaultOrgID)
+	if err != nil {
+		t.Fatalf("open store: %v", err)
+	}
+	defer store.Close()
+
+	saved := ResourceOperatorState{
+		CanonicalID:          "docker:host-1/ct-nginx",
+		IntentionallyOffline: true,
+		SetAt:                time.Date(2026, 7, 20, 10, 0, 0, 0, time.UTC),
+		SetBy:                "operator:richard",
+	}
+	if err := store.SetResourceOperatorState(saved); err != nil {
+		t.Fatalf("set: %v", err)
+	}
+
+	got, found, err := store.GetResourceOperatorState(saved.CanonicalID)
+	if err != nil {
+		t.Fatalf("get: %v", err)
+	}
+	if !found {
+		t.Fatal("expected entry after set")
+	}
+	if got.AutoRemediationPolicy.CapabilityNames == nil {
+		t.Fatal("CapabilityNames must be normalized when the policy column is NULL")
+	}
+
+	encoded, err := json.Marshal(got)
+	if err != nil {
+		t.Fatalf("marshal: %v", err)
+	}
+	if strings.Contains(string(encoded), `"capabilityNames":null`) ||
+		!strings.Contains(string(encoded), `"capabilityNames":[]`) {
+		t.Fatalf("wire payload must contain an empty capabilityNames list, got: %s", encoded)
+	}
+}

internal/unifiedresources/store.go

@@ -2931,8 +2931,11 @@ func scanResourceOperatorState(scanner resourceOperatorStateScanner) (ResourceOp
 		if err := json.Unmarshal([]byte(autoPolicyJSON.String), &state.AutoRemediationPolicy); err != nil {
 			return ResourceOperatorState{}, fmt.Errorf("unmarshal auto remediation policy: %w", err)
 		}
-		state.AutoRemediationPolicy = NormalizeAutoRemediationPolicy(state.AutoRemediationPolicy)
 	}
+	// Normalize even when the column was NULL (row saved with an empty
+	// policy) so CapabilityNames is never nil and the wire carries [],
+	// not null (issue #1621).
+	state.AutoRemediationPolicy = NormalizeAutoRemediationPolicy(state.AutoRemediationPolicy)
 	if startAt.Valid {
 		t := startAt.Time
 		state.MaintenanceStartAt = &t