Judge source merges by the ingest's stale thresholds (#2717)

Stacked on #2715 (its commit is the parent of this one); either can land
first.

## Problem

`mergeInto` passed nil thresholds to `mergeMetrics`, so every source
merge into an existing row (snapshot and record ingest alike) judged
metric freshness by the registry defaults, while the stale pass and
manual links used the monitor's configured thresholds. Two visible
results:

- A Proxmox node polled every two minutes (four-minute threshold) merged
with its auto-linked agent showed the silent agent's CPU and memory,
while the stale pass called the node's poll current.
- A supplemental refresh (`MonitorAdapter.PopulateSupplementalRecords`,
and the read-state overlay) of a linked vSphere VM kept the in-guest
agent's memory over vSphere's current reading, where a full rebuild from
the same records shows vSphere's.

## Change

- Snapshot and record ingest hold the caller's thresholds on the
registry while they run (the same per-pass pattern as the guest loop's
agent index), and `mergeInto` passes them to `mergeMetrics`. The
snapshot helpers that reach `mergeInto` keep their signatures.
- The stale pass and the metric merge share one threshold rule,
`sourceStaleThreshold`, so the merge also honours a sighting's expected
update interval (slow Proxmox disk inventory), as the stale pass already
did.
- Outside mock mode the monitor configures only Proxmox, PBS and PMG, so
those are the production sources this changes; mock mode also configures
TrueNAS, vSphere and availability.

## Proof

- `TestMonitorAdapterSourceMergesUseConfiguredStaleThresholds`: rebuild,
supplemental refresh and overlay of a linked vSphere VM, plus a node
with an auto-linked agent, through
`NewMonitorAdapterWithStaleThresholds`.
- `TestSourceMergesJudgeFreshnessByTheIngestStaleThresholds`: snapshot
node+agent merge with and without thresholds.
- `TestMetricMergeFreshnessMatchesTheStalePass`: the merge and the stale
pass agree for default, configured and expected-interval sightings.
- Negative control with #2715's `registry.go`: the supplemental,
overlay, node+agent, snapshot and expected-interval cases fail; the
rebuild case and the defaults case pass on both.
- `go test ./internal/unifiedresources/... ./internal/monitoring/...
./internal/api/resourceapi/... ./internal/mock/...` on pulse-dev.

docs/release-control/v6/internal/subsystems/monitoring.md

@@ -2699,7 +2699,18 @@ truthfulness, not native thaw, containing-release or workload acceptance.
     ingestion so canonical host IDs stay stable across restarts (see the
     unified-resources contract's durable identity-pin obligation). Rebuild
     paths added to the adapter must keep that persistence step; ephemeral
-    snapshot-bridge adapters stay read-only.
+    snapshot-bridge adapters stay read-only. The rebuild, the live
+    supplemental refresh and the read-state overlay ingest records with the
+    adapter's configured stale thresholds, because record ingest joins
+    operator links and the freshness gate of every metric merge, a link's or
+    a source's into an existing row, reads them (unified-resources contract,
+    "Operator links reach record-ingested resources"). Regression
+    coverage:
+    `TestMonitorAdapterJoinsLinkedRecordsWithConfiguredStaleThresholds` and
+    `TestMonitorAdapterSourceMergesUseConfiguredStaleThresholds` in
+    `internal/unifiedresources/monitor_adapter_read_state_test.go` and
+    `TestManualLinkToSupplementalGuestHoldsWithAndWithoutContinuity` in
+    `internal/monitoring/issue1913_host_continuity_test.go`.
 
 11. The TrueNAS provider projects pools with `Storage.Topology` fixed to
     `pool` and the ZFS data vdev layout in `Storage.VDevLayout`. The

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

@@ -1609,7 +1609,12 @@ Source freshness is a separate unified-resource status input, not a fixed
 global health timer. Snapshot rebuilds, resource seeding, supplemental record
 overlays, and cloned read-state overlays must preserve the monitoring-provided
 stale thresholds for Proxmox, PBS, and PMG sources so resources do not flap to
-warning/degraded between successful configured poll cycles.
+warning/degraded between successful configured poll cycles. A source merge
+into an existing row judges metric freshness by the ingest's thresholds and
+the stale pass's own threshold rule, so the merge's freshness gate never
+calls a source current that the stale pass marks stale, or the reverse. The
+gate decides only where no stronger rule does: a hypervisor-managed guest
+keeps its platform CPU even while that source is stale.
 
 Service-discovery readiness is a unified-resource payload contract, not a
 drawer-local decoration. Resource list/detail payloads that expose a
@@ -2813,7 +2818,9 @@ application resource-provider or WebSocket lifecycle.
     ID and drops the superseded pin row (`ApplyCanonicalIDSuccessions`
     in `internal/unifiedresources/canonical_id_succession.go`).
     Successions never fire while the old ID still belongs to a live
-    resource, never follow a contradicting machine key (a reinstalled
+    resource or to one a manual link folded into its primary (see
+    "Operator links reach record-ingested resources"), never follow a
+    contradicting machine key (a reinstalled
     machine mints fresh, it does not absorb the old host's operator
     state), and never rewrite change-journal rows. Known limitation:
     machine-keyless hosts whose history already merged under a collapsed
@@ -2894,8 +2901,9 @@ application resource-provider or WebSocket lifecycle.
     `ResourceRegistry.IngestRecords` applies the same
     `ApplyCanonicalIDSuccessions` semantics as pin-driven successions
     (operator state and action audits re-key, the superseded pin row
-    drops, never while the old ID belongs to a live resource, journal
-    rows never rewritten). Regression coverage:
+    drops, never while the old ID belongs to a live resource or one a
+    manual link folded into its primary, journal rows never rewritten).
+    Regression coverage:
     `TestRegistryIngestRecordsKeepsSameHostnameSystemsDistinct` /
     `TestIngestRecordsSucceedLegacyHostnameScopedCanonicalIDs` /
     `TestIngestRecordsDoNotCompleteTrueNASIdentityFromPins` in
@@ -6596,6 +6604,75 @@ full heading. This shared drawer rule applies across resource types. The 390px
 browser qualification checks heading fit and keyboard expansion; no data model,
 authorisation, action dispatch or resource admission contract changes.
 
+### Operator links reach record-ingested resources
+
+Every registry ingest path applies the operator's manual links: snapshot
+ingest, resource ingest and record ingest (`IngestRecords`,
+`IngestRecordsWithStaleThresholds` and the read-state overlay's record pass).
+Either side of a link can arrive only as a supplemental record, such as a
+vSphere or TrueNAS VM with a Pulse agent inside, or an agent on a TrueNAS host,
+and the monitor's rebuild ingests those records after the snapshot. While only
+snapshot and resource ingest applied links, the rebuilt registry behind the
+websocket broadcast and `/api/state` kept both rows. The resources API, which
+seeds through resource ingest, showed them merged, and so did the published
+inventory whenever an unrelated absent machine made continuity hydration
+clone the registry. Mock mode builds its unified view from fixtures without
+the link store, so links do not reach the mock broadcast; that boundary is
+separate.
+
+- The monitor adapter passes its configured stale thresholds to record
+  ingest, so a link joined there judges each side's metrics the way a
+  snapshot-time link does. Ordinary source merges into an existing row use
+  the same thresholds on every ingest path: snapshot and record ingest hold
+  the caller's thresholds while they run, and `mergeInto` judges each
+  metric's freshness with them through `sourceStaleThreshold`, the rule the
+  stale pass reads. Before this, those merges used the defaults, so a
+  Proxmox node polled every two minutes showed a silent auto-linked agent's
+  readings while the stale pass called the node's poll current, and a
+  supplemental refresh of a linked vSphere VM kept the agent's memory over
+  vSphere's current reading. Outside mock mode the monitor configures only
+  Proxmox, PBS and PMG; mock mode also configures TrueNAS, vSphere and
+  availability. Regression coverage:
+  `TestMonitorAdapterSourceMergesUseConfiguredStaleThresholds` (rebuild,
+  supplemental refresh, overlay, node with auto-linked agent),
+  `TestSourceMergesJudgeFreshnessByTheIngestStaleThresholds` and
+  `TestMetricMergeFreshnessMatchesTheStalePass`.
+- Continuity records are the one record ingest that never joins links. A
+  saved enrollment only fills an absent machine, so it stays its own row
+  instead of lending its linked guest an offline verdict and an old agent
+  payload.
+- A link merge keeps the TrueNAS and vSphere payloads the primary lacks, so
+  an agent chosen as primary over a TrueNAS system still carries the
+  system's facet, and the system's pools re-parent to the merged row.
+- The registry remembers each resource a manual link folded into its
+  primary. A folded resource is still observed, so neither pin-driven nor
+  record-declared succession may succeed its ID. An agent-type primary
+  carries the folded resource's machine keys and is pinned with them, and
+  succeeding the folded ID re-keyed the link onto the primary itself, which
+  split the pair on the next rebuild. That already broke snapshot links whose
+  primary is a machine-keyless agent-type row, such as a Proxmox node.
+- Fold history is registry-local: resource seeding and overlay clones do not
+  carry it. A reference to the folded resource's canonical ID, such as an
+  availability check configured against an agent later linked into its
+  guest, does not resolve to the primary, so the check loses its target in
+  the broadcast as it already did in the resources API.
+
+`registry_merge_policy_test.go` pins the record links in both directions
+through two rebuilds and the live supplemental refresh, and checks that a
+registry seeded from the monitor's listing through resource ingest, as the
+resources API seeds, lists the same rows
+(`TestManualLinksJoinRecordIngestedResourcesInMonitorRebuild`). The same file
+pins the node case (`TestManualLinkKeepsItsEndpointsThroughIdentityPinPersistence`),
+record-declared succession (`TestRecordDeclaredSuccessionSparesALinkFoldedResource`)
+and both payloads (`TestManualLinkKeepsProviderPayloadsThePrimaryLacks`).
+`monitor_adapter_read_state_test.go`
+(`TestMonitorAdapterJoinsLinkedRecordsWithConfiguredStaleThresholds`) pins the
+thresholds on the rebuild, supplemental and overlay paths, and
+`internal/monitoring/issue1913_host_continuity_test.go`
+(`TestManualLinkToSupplementalGuestHoldsWithAndWithoutContinuity`) pins the
+published inventory with no continuity, an unrelated absent machine and the
+linked agent's own saved enrollment.
+
 ### Provider link network corroboration
 
 An existing reciprocal host/provider link remains authoritative, including a

internal/monitoring/issue1913_host_continuity_test.go

@@ -89,3 +89,120 @@ func TestHostContinuityPreservesLiveAgentAfterReenrollment(t *testing.T) {
 		})
 	}
 }
+
+// Continuity hydration clones the registry through resource ingest, which
+// applies operator links, while the rebuild ingests supplemental records after
+// the snapshot. A link to a record-sourced guest (an agent inside a vSphere VM)
+// must hold in the published inventory either way, or the state the websocket
+// broadcasts and the resources API seeds from flips with an unrelated absent
+// machine. A saved enrollment only fills an absent machine: when the linked
+// agent itself is gone, its continuity row stays separate and the guest keeps
+// what vSphere reports.
+func TestManualLinkToSupplementalGuestHoldsWithAndWithoutContinuity(t *testing.T) {
+	now := time.Now().UTC()
+	guestAgent := models.Host{
+		ID: "host-app-guest", MachineID: "machine-app-guest", Hostname: "app-guest",
+		Platform: "linux", Status: "online", LastSeen: now, IntervalSeconds: 30,
+	}
+	vmRecords := []unifiedresources.IngestRecord{{
+		SourceID: "vc-1:vm:vm-42",
+		Resource: unifiedresources.Resource{
+			Type:       unifiedresources.ResourceTypeVM,
+			Technology: "vmware",
+			Name:       "app-guest",
+			Status:     unifiedresources.StatusOnline,
+			LastSeen:   now,
+			VMware:     &unifiedresources.VMwareData{ConnectionID: "vc-1", ManagedObjectID: "vm-42", EntityType: "vm"},
+		},
+	}}
+
+	unlinked := unifiedresources.NewMonitorAdapter(unifiedresources.NewRegistry(nil))
+	unlinked.PopulateSnapshotAndSupplemental(
+		models.StateSnapshot{Hosts: []models.Host{guestAgent}, LastUpdate: now},
+		map[unifiedresources.DataSource][]unifiedresources.IngestRecord{unifiedresources.SourceVMware: vmRecords},
+	)
+	if len(unlinked.VMs()) != 1 || len(unlinked.Hosts()) != 1 {
+		t.Fatalf("unlinked estate = %d VMs, %d agents, want one of each", len(unlinked.VMs()), len(unlinked.Hosts()))
+	}
+	vmID, agentID := unlinked.VMs()[0].ID(), unlinked.Hosts()[0].ID()
+
+	for _, tc := range []struct {
+		name       string
+		reporting  bool
+		continuity []config.HostContinuityEntry
+	}{
+		{name: "no-continuity", reporting: true},
+		{
+			name:      "absent-machine-continuity",
+			reporting: true,
+			continuity: []config.HostContinuityEntry{{
+				HostID: "absent-machine", MachineID: "other-machine", Hostname: "other-host",
+				Platform: "linux", LastSeen: now.Add(-time.Hour),
+			}},
+		},
+		{
+			name: "linked-agent-continuity-only",
+			continuity: []config.HostContinuityEntry{{
+				HostID: guestAgent.ID, MachineID: guestAgent.MachineID, Hostname: guestAgent.Hostname,
+				Platform: "linux", LastSeen: now.Add(-10 * time.Minute),
+			}},
+		},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			links := unifiedresources.NewMemoryStore()
+			if err := links.AddLink(unifiedresources.ResourceLink{ResourceA: vmID, ResourceB: agentID, PrimaryID: vmID}); err != nil {
+				t.Fatalf("add link: %v", err)
+			}
+			state := models.NewState()
+			if tc.reporting {
+				state.Hosts = []models.Host{guestAgent}
+			}
+			m := &Monitor{state: state}
+			if len(tc.continuity) > 0 {
+				continuity := config.NewHostContinuityStore(t.TempDir(), nil)
+				for _, entry := range tc.continuity {
+					if err := continuity.Upsert(entry); err != nil {
+						t.Fatal(err)
+					}
+				}
+				m.hostContinuityStore = continuity
+			}
+			m.SetResourceStore(unifiedresources.NewMonitorAdapter(unifiedresources.NewRegistry(links)))
+			m.SetSupplementalRecordsProvider(unifiedresources.SourceVMware, &testSupplementalInventoryReadinessProvider{
+				source:  unifiedresources.SourceVMware,
+				records: vmRecords,
+				settled: true,
+				readyAt: now,
+			})
+
+			resources, _ := m.UnifiedResourceSnapshot()
+			var guest, agent *unifiedresources.Resource
+			for i := range resources {
+				switch resources[i].ID {
+				case agentID:
+					agent = &resources[i]
+				case vmID:
+					guest = &resources[i]
+				}
+			}
+			if guest == nil {
+				t.Fatalf("linked VM %s missing from published inventory", vmID)
+			}
+			if tc.reporting {
+				if agent != nil {
+					t.Fatalf("linked agent %s still published standalone", agentID)
+				}
+				if guest.VMware == nil || guest.Agent == nil {
+					t.Fatalf("published VM vmware=%t agent=%t, want the VM carrying its agent", guest.VMware != nil, guest.Agent != nil)
+				}
+				return
+			}
+			if agent == nil || agent.Status != unifiedresources.StatusOffline {
+				t.Fatalf("saved enrollment row = %+v, want its own offline row", agent)
+			}
+			if guest.Agent != nil || guest.Status != unifiedresources.StatusOnline {
+				t.Fatalf("published VM agent=%t status=%s, want vSphere's online VM untouched by the saved enrollment", guest.Agent != nil, guest.Status)
+			}
+		})
+	}
+}

internal/unifiedresources/canonical_id_pins.go

@@ -337,8 +337,9 @@ func (index *identityPinIndex) successionsFor(pin ResourceIdentityPin) []Canonic
 // the new canonical ID before the pin write, so operator intent
 // (never-auto-remediate, maintenance windows) and action-audit history
 // survive the era change. Successions are skipped while the old canonical ID
-// still belongs to a live resource: a genuinely short-named host must not
-// have its rows stolen by an FQDN sibling. Change-journal rows are never
+// still belongs to a live resource, or to one a manual link folded into
+// another: a genuinely short-named host must not have its rows stolen by an
+// FQDN sibling. Change-journal rows are never
 // rewritten; EraIDs merges those at read time.
 func (rr *ResourceRegistry) PersistIdentityPins() {
 	if rr.store == nil {
@@ -358,7 +359,11 @@ func (rr *ResourceRegistry) PersistIdentityPins() {
 		}
 		pins = append(pins, pin)
 		for _, succession := range rr.identityPins.successionsFor(pin) {
-			if _, live := rr.resources[succession.OldCanonicalID]; live {
+			// A primary that absorbed a linked resource also carries its
+			// machine keys. That resource is still observed, and succeeding
+			// it would re-key the link onto the primary itself, splitting
+			// the pair again on the next rebuild.
+			if rr.canonicalIDObservedLocked(succession.OldCanonicalID) {
 				continue
 			}
 			successions = append(successions, succession)

internal/unifiedresources/monitor_adapter.go

@@ -135,7 +135,7 @@ func readStateWithRecords(readState ReadState, source DataSource, records []Inge
 	cloned := NewRegistry(registry.store)
 	thresholds := adapter.currentStaleThresholds()
 	cloned.IngestResourcesWithStaleThresholds(registry.List(), thresholds)
-	cloned.ingestRecords(source, records, onlyMissing)
+	cloned.ingestRecords(source, records, onlyMissing, thresholds)
 	overlay := NewMonitorAdapterWithStaleThresholds(cloned, thresholds)
 	if keyed {
 		adapter.overlays.store(registry, rebuiltAt, key, now, overlay)
@@ -450,7 +450,7 @@ func (a *MonitorAdapter) replaceRegistryLocked(snapshot models.StateSnapshot, re
 		if len(records) == 0 || strings.TrimSpace(string(source)) == "" {
 			continue
 		}
-		rebuilt.IngestRecords(source, records)
+		rebuilt.IngestRecordsWithStaleThresholds(source, records, staleThresholds)
 	}
 	// IngestSnapshot runs its stale pass before the record sources above are
 	// ingested, so record-sourced resources would otherwise keep their
@@ -665,7 +665,7 @@ func (a *MonitorAdapter) PopulateSupplementalRecords(source DataSource, records
 		return
 	}
 	before := registry.List()
-	registry.IngestRecords(source, records)
+	registry.IngestRecordsWithStaleThresholds(source, records, a.currentStaleThresholds())
 	rebuiltAt := time.Now().UTC()
 	recordRegistryChanges(registry.store, before, registry.List(), rebuiltAt, nil, SourcePlatformEvent, changeSourceAdapterForDataSource(source))
 	registry.PersistIdentityPins()

internal/unifiedresources/monitor_adapter_read_state_test.go

@@ -400,6 +400,234 @@ func TestReadStateWithRecordsPreservesConfiguredStaleThresholds(t *testing.T) {
 	}
 }
 
+// A link whose side arrives as a record joins during record ingest, so it
+// must judge freshness by the adapter's thresholds as a snapshot-time link
+// does. Under a five-minute vSphere threshold a VM polled three minutes ago
+// is current, and its own memory reading outranks its in-guest agent's.
+func TestMonitorAdapterJoinsLinkedRecordsWithConfiguredStaleThresholds(t *testing.T) {
+	now := time.Now().UTC()
+	snapshot := models.StateSnapshot{
+		Hosts: []models.Host{{
+			ID:        "host-app-guest",
+			Hostname:  "app-guest",
+			MachineID: "machine-app-guest",
+			Status:    "online",
+			LastSeen:  now,
+			Memory:    models.Memory{Total: 8 << 30, Used: 6 << 30, Free: 2 << 30, Usage: 75},
+		}},
+		LastUpdate: now,
+	}
+	vmRecords := []IngestRecord{{
+		SourceID: "vc-1:vm:vm-42",
+		Resource: Resource{
+			Type:       ResourceTypeVM,
+			Technology: "vmware",
+			Name:       "app-guest",
+			Status:     StatusOnline,
+			LastSeen:   now.Add(-3 * time.Minute),
+			Metrics:    &ResourceMetrics{Memory: &MetricValue{Percent: 40, Source: SourceVMware}},
+			VMware:     &VMwareData{ConnectionID: "vc-1", ManagedObjectID: "vm-42", EntityType: "vm"},
+		},
+	}}
+	thresholds := map[DataSource]time.Duration{SourceVMware: 5 * time.Minute}
+
+	unlinked := NewMonitorAdapter(NewRegistry(nil))
+	unlinked.PopulateSnapshotAndSupplemental(snapshot, map[DataSource][]IngestRecord{SourceVMware: vmRecords})
+	vms, hosts := unlinked.VMs(), unlinked.Hosts()
+	if len(vms) != 1 || len(hosts) != 1 {
+		t.Fatalf("unlinked estate = %d VMs, %d agents, want one of each", len(vms), len(hosts))
+	}
+	link := ResourceLink{ResourceA: vms[0].ID(), ResourceB: hosts[0].ID(), PrimaryID: hosts[0].ID()}
+
+	for _, path := range []struct {
+		name string
+		read func(adapter *MonitorAdapter) ReadState
+	}{
+		{"rebuild", func(adapter *MonitorAdapter) ReadState {
+			adapter.PopulateSnapshotAndSupplemental(snapshot, map[DataSource][]IngestRecord{SourceVMware: vmRecords})
+			return adapter
+		}},
+		{"supplemental", func(adapter *MonitorAdapter) ReadState {
+			adapter.PopulateFromSnapshot(snapshot)
+			adapter.PopulateSupplementalRecords(SourceVMware, vmRecords)
+			return adapter
+		}},
+		{"overlay", func(adapter *MonitorAdapter) ReadState {
+			adapter.PopulateFromSnapshot(snapshot)
+			return ReadStateWithRecords(adapter, SourceVMware, vmRecords)
+		}},
+	} {
+		t.Run(path.name, func(t *testing.T) {
+			store := NewMemoryStore()
+			if err := store.AddLink(link); err != nil {
+				t.Fatalf("add link: %v", err)
+			}
+			readState := path.read(NewMonitorAdapterWithStaleThresholds(NewRegistry(store), thresholds))
+
+			if hosts := readState.Hosts(); len(hosts) != 0 {
+				t.Fatalf("linked agent still listed standalone: %s", hosts[0].ID())
+			}
+			vms := readState.VMs()
+			if len(vms) != 1 || vms[0].ID() != link.ResourceA {
+				t.Fatalf("VMs = %d, want the linked VM %s", len(vms), link.ResourceA)
+			}
+			if vms[0].r.Metrics == nil || vms[0].r.Metrics.Memory == nil {
+				t.Fatal("linked VM lost its memory reading")
+			}
+			if memory := vms[0].r.Metrics.Memory; memory.Source != SourceVMware || memory.Percent != 40 {
+				t.Fatalf("linked VM memory = %.0f%% from %s, want vSphere's current 40%%", memory.Percent, memory.Source)
+			}
+		})
+	}
+}
+
+// Every source merge into an existing row judges metric freshness by the
+// adapter's configured thresholds, as the rebuild's links and stale pass do.
+// A linked vSphere VM first reports no memory, so its in-guest agent supplies
+// it; a refresh then brings vSphere memory observed three minutes ago, current
+// under a five-minute threshold. The live supplemental refresh and the
+// read-state overlay merge that record into the existing row, and must show
+// vSphere's memory as a rebuild from the same records does. A Proxmox node
+// polled every two minutes is likewise current at ninety seconds, so the
+// rebuild keeps its readings over those of its auto-linked agent, silent for
+// seventy-five seconds.
+func TestMonitorAdapterSourceMergesUseConfiguredStaleThresholds(t *testing.T) {
+	now := time.Now().UTC()
+	thresholds := map[DataSource]time.Duration{SourceProxmox: 4 * time.Minute, SourceVMware: 5 * time.Minute}
+
+	t.Run("linked vSphere VM refresh", func(t *testing.T) {
+		snapshot := models.StateSnapshot{
+			Hosts: []models.Host{{
+				ID:        "host-app-guest",
+				Hostname:  "app-guest",
+				MachineID: "machine-app-guest",
+				Status:    "online",
+				LastSeen:  now,
+				Memory:    models.Memory{Total: 8 << 30, Used: 6 << 30, Free: 2 << 30, Usage: 75},
+			}},
+			LastUpdate: now,
+		}
+		vmRecord := func(lastSeen time.Time, metrics *ResourceMetrics) []IngestRecord {
+			return []IngestRecord{{
+				SourceID: "vc-1:vm:vm-42",
+				Resource: Resource{
+					Type:       ResourceTypeVM,
+					Technology: "vmware",
+					Name:       "app-guest",
+					Status:     StatusOnline,
+					LastSeen:   lastSeen,
+					Metrics:    metrics,
+					VMware:     &VMwareData{ConnectionID: "vc-1", ManagedObjectID: "vm-42", EntityType: "vm"},
+				},
+			}}
+		}
+		initial := vmRecord(now.Add(-4*time.Minute), &ResourceMetrics{CPU: &MetricValue{Percent: 10, Source: SourceVMware}})
+		refreshedAt := now.Add(-3 * time.Minute)
+		refreshed := vmRecord(refreshedAt, &ResourceMetrics{
+			CPU:    &MetricValue{Percent: 10, Source: SourceVMware},
+			Memory: &MetricValue{Percent: 40, Source: SourceVMware},
+		})
+
+		unlinked := NewMonitorAdapter(NewRegistry(nil))
+		unlinked.PopulateSnapshotAndSupplemental(snapshot, map[DataSource][]IngestRecord{SourceVMware: initial})
+		vms, hosts := unlinked.VMs(), unlinked.Hosts()
+		if len(vms) != 1 || len(hosts) != 1 {
+			t.Fatalf("unlinked estate = %d VMs, %d agents, want one of each", len(vms), len(hosts))
+		}
+		link := ResourceLink{ResourceA: vms[0].ID(), ResourceB: hosts[0].ID(), PrimaryID: hosts[0].ID()}
+
+		for _, path := range []struct {
+			name string
+			read func(adapter *MonitorAdapter) ReadState
+		}{
+			{"rebuild", func(adapter *MonitorAdapter) ReadState {
+				adapter.PopulateSnapshotAndSupplemental(snapshot, map[DataSource][]IngestRecord{SourceVMware: refreshed})
+				return adapter
+			}},
+			{"supplemental", func(adapter *MonitorAdapter) ReadState {
+				adapter.PopulateSnapshotAndSupplemental(snapshot, map[DataSource][]IngestRecord{SourceVMware: initial})
+				adapter.PopulateSupplementalRecords(SourceVMware, refreshed)
+				return adapter
+			}},
+			{"overlay", func(adapter *MonitorAdapter) ReadState {
+				adapter.PopulateSnapshotAndSupplemental(snapshot, map[DataSource][]IngestRecord{SourceVMware: initial})
+				return ReadStateWithRecords(adapter, SourceVMware, refreshed)
+			}},
+		} {
+			t.Run(path.name, func(t *testing.T) {
+				store := NewMemoryStore()
+				if err := store.AddLink(link); err != nil {
+					t.Fatalf("add link: %v", err)
+				}
+				readState := path.read(NewMonitorAdapterWithStaleThresholds(NewRegistry(store), thresholds))
+
+				vms := readState.VMs()
+				if len(vms) != 1 || vms[0].ID() != link.ResourceA || len(readState.Hosts()) != 0 {
+					t.Fatalf("estate = %d VMs, %d standalone agents, want only the linked VM %s", len(vms), len(readState.Hosts()), link.ResourceA)
+				}
+				if got := vms[0].r.SourceStatus[SourceVMware]; got.Status != "online" || !got.LastSeen.Equal(refreshedAt) {
+					t.Fatalf("vSphere sighting = %+v, want the refresh, current under the configured threshold", got)
+				}
+				if vms[0].r.Metrics == nil || vms[0].r.Metrics.Memory == nil {
+					t.Fatal("linked VM lost its memory reading")
+				}
+				if memory := vms[0].r.Metrics.Memory; memory.Source != SourceVMware || memory.Percent != 40 {
+					t.Fatalf("linked VM memory = %.0f%% from %s, want vSphere's current 40%%", memory.Percent, memory.Source)
+				}
+			})
+		}
+	})
+
+	t.Run("node with auto-linked agent", func(t *testing.T) {
+		adapter := NewMonitorAdapterWithStaleThresholds(NewRegistry(nil), thresholds)
+		adapter.PopulateFromSnapshot(models.StateSnapshot{
+			Nodes: []models.Node{{
+				ID:            "homelab-pve1",
+				Name:          "pve1",
+				Instance:      "homelab",
+				Status:        "online",
+				LinkedAgentID: "host-pve1",
+				CPU:           0.12,
+				Memory:        models.Memory{Total: 64 << 30, Used: 16 << 30, Free: 48 << 30, Usage: 25},
+				LastSeen:      now.Add(-90 * time.Second),
+			}},
+			Hosts: []models.Host{{
+				ID:              "host-pve1",
+				MachineID:       "machine-pve1",
+				Hostname:        "pve1",
+				LinkedNodeID:    "homelab-pve1",
+				Status:          "online",
+				CPUUsage:        80,
+				Memory:          models.Memory{Total: 64 << 30, Used: 48 << 30, Free: 16 << 30, Usage: 75},
+				LastSeen:        now.Add(-75 * time.Second),
+				IntervalSeconds: 30,
+			}},
+			LastUpdate: now,
+		})
+
+		nodes := adapter.Nodes()
+		if len(nodes) != 1 {
+			t.Fatalf("nodes = %d, want the node merged with its agent", len(nodes))
+		}
+		node := nodes[0].r
+		if poll := node.SourceStatus[SourceProxmox]; poll.Status != "online" || !poll.LastSeen.Equal(now.Add(-90*time.Second)) {
+			t.Fatalf("Proxmox sighting = %+v, want the current poll from ninety seconds ago", poll)
+		}
+		if got := node.SourceStatus[SourceAgent].Status; got != "stale" {
+			t.Fatalf("agent sighting = %q, want stale", got)
+		}
+		if node.Metrics == nil || node.Metrics.CPU == nil || node.Metrics.Memory == nil {
+			t.Fatalf("merged node lost its readings: %+v", node.Metrics)
+		}
+		if cpu := node.Metrics.CPU; cpu.Source != SourceProxmox || cpu.Percent != 12 {
+			t.Fatalf("node CPU = %.0f%% from %s, want the current poll's 12%%", cpu.Percent, cpu.Source)
+		}
+		if memory := node.Metrics.Memory; memory.Source != SourceProxmox || memory.Percent != 25 {
+			t.Fatalf("node memory = %.0f%% from %s, want the current poll's 25%%", memory.Percent, memory.Source)
+		}
+	})
+}
+
 func TestHostContinuityCannotOverwriteCurrentResource(t *testing.T) {
 	now := time.Now().UTC()
 	for _, mode := range []string{"machine-identity", "source-identity", "future-timestamp", "provider-only"} {

internal/unifiedresources/registry.go

@@ -95,6 +95,10 @@ type ResourceRegistry struct {
 	exclusions   map[string]struct{}
 	identityPins *identityPinIndex
 	pbsBackups   []models.PBSBackup
+	// linkMergedIDs holds the resources a manual link folded into their
+	// primary. They are still observed, so identity succession treats them
+	// as live.
+	linkMergedIDs map[string]struct{}
 	// supersededIndex maps record-declared retired canonical IDs to the live
 	// resource that superseded them, so references persisted under a retired
 	// ID (availability links, API reads) keep resolving. An empty value marks
@@ -148,6 +152,13 @@ type ResourceRegistry struct {
 	// what the full scan would see; scores stay live through the shared
 	// pointers.
 	agentNodeScanIndex map[string][]agentNodeCandidate
+
+	// ingestStaleThresholds holds the caller-owned freshness thresholds of
+	// the snapshot or record ingest in progress, so each source merge into an
+	// existing row judges metric freshness as that ingest's manual links and
+	// stale pass do. nil outside an ingest and for callers without
+	// thresholds, where all three use the defaults.
+	ingestStaleThresholds map[DataSource]time.Duration
 }
 
 // agentNodeCandidate pairs a resources-map key with its entry so the indexed
@@ -243,6 +254,10 @@ func (rr *ResourceRegistry) IngestSnapshotWithStaleThresholds(snapshot models.St
 }
 
 func (rr *ResourceRegistry) ingestSnapshot(snapshot models.StateSnapshot, thresholds map[DataSource]time.Duration) {
+	rr.mu.Lock()
+	rr.ingestStaleThresholds = thresholds
+	rr.mu.Unlock()
+
 	hostByID := make(map[string]*models.Host, len(snapshot.Hosts))
 	for i := range snapshot.Hosts {
 		host := snapshot.Hosts[i]
@@ -606,15 +621,26 @@ func (rr *ResourceRegistry) ingestSnapshot(snapshot models.StateSnapshot, thresh
 	rr.refreshCanonicalIdentitiesLocked()
 	rr.markStaleLocked(time.Now().UTC(), thresholds)
 	rr.invalidateViewsLocked()
+	rr.ingestStaleThresholds = nil
 	rr.mu.Unlock()
 }
 
 // IngestRecords ingests normalized records for a single source.
 func (rr *ResourceRegistry) IngestRecords(source DataSource, records []IngestRecord) {
-	rr.ingestRecords(source, records, false)
+	rr.ingestRecords(source, records, false, nil)
 }
 
-func (rr *ResourceRegistry) ingestRecords(source DataSource, records []IngestRecord, onlyMissing bool) {
+// IngestRecordsWithStaleThresholds ingests normalized records for a single
+// source and merges operator-linked resources with caller-owned thresholds.
+func (rr *ResourceRegistry) IngestRecordsWithStaleThresholds(source DataSource, records []IngestRecord, thresholds map[DataSource]time.Duration) {
+	rr.ingestRecords(source, records, false, thresholds)
+}
+
+func (rr *ResourceRegistry) ingestRecords(source DataSource, records []IngestRecord, onlyMissing bool, thresholds map[DataSource]time.Duration) {
+	rr.mu.Lock()
+	rr.ingestStaleThresholds = thresholds
+	rr.mu.Unlock()
+
 	var successions []CanonicalIDSuccession
 	supersededSeen := make(map[string]struct{})
 	for _, record := range records {
@@ -649,6 +675,16 @@ func (rr *ResourceRegistry) ingestRecords(source DataSource, records []IngestRec
 	rr.applyRecordSuccessions(successions)
 
 	rr.mu.Lock()
+	// Either side of an operator link can arrive as a record: a vSphere or
+	// TrueNAS VM with a Pulse agent inside, or an agent on a TrueNAS host.
+	// Snapshot and resource ingest apply the links, so record ingest must too,
+	// or the monitor's rebuilt registry keeps both rows while REST, which
+	// seeds from that registry, shows them merged. Continuity records only
+	// fill absent machines: a saved enrollment joined to its linked guest
+	// would lend the guest its offline verdict and old agent payload.
+	if !onlyMissing {
+		rr.applyManualLinks(thresholds)
+	}
 	rr.refreshStorageConsumersLocked()
 	rr.refreshPBSRollupsLocked()
 	rr.refreshStoragePostureLocked()
@@ -657,6 +693,7 @@ func (rr *ResourceRegistry) ingestRecords(source DataSource, records []IngestRec
 	rr.refreshProxmoxSensorSetupLocked()
 	rr.refreshCanonicalIdentitiesLocked()
 	rr.invalidateViewsLocked()
+	rr.ingestStaleThresholds = nil
 	rr.mu.Unlock()
 }
 
@@ -746,12 +783,25 @@ func (rr *ResourceRegistry) proxmoxGuestResourceIDForSourceRefLocked(ref string)
 	return uniqueID
 }
 
+// canonicalIDObservedLocked reports whether a canonical ID still names an
+// observed resource: a live row, or one a manual link folded into its primary.
+// Succeeding an observed ID would hand its operator rows, and the link that
+// folded it, to another resource.
+func (rr *ResourceRegistry) canonicalIDObservedLocked(id string) bool {
+	if _, live := rr.resources[id]; live {
+		return true
+	}
+	_, folded := rr.linkMergedIDs[id]
+	return folded
+}
+
 // applyRecordSuccessions re-keys operator-owned store rows from canonical IDs
 // that ingested records declared superseded (IngestRecord.SupersededCanonicalIDs)
 // onto the records' current canonical IDs. Mirrors the guards of pin-driven
-// successions in PersistIdentityPins: a superseded ID still held by a live
-// resource is skipped so a genuinely distinct sibling never has its rows
-// stolen. Re-runs are cheap no-ops once the old rows are gone.
+// successions in PersistIdentityPins: a superseded ID still observed, live or
+// folded into another resource by a manual link, is skipped so a genuinely
+// distinct sibling never has its rows stolen. Re-runs are cheap no-ops once
+// the old rows are gone.
 func (rr *ResourceRegistry) applyRecordSuccessions(successions []CanonicalIDSuccession) {
 	if len(successions) == 0 || rr.store == nil {
 		return
@@ -764,7 +814,7 @@ func (rr *ResourceRegistry) applyRecordSuccessions(successions []CanonicalIDSucc
 	rr.mu.RLock()
 	kept := make([]CanonicalIDSuccession, 0, len(successions))
 	for _, succession := range successions {
-		if _, live := rr.resources[succession.OldCanonicalID]; live {
+		if rr.canonicalIDObservedLocked(succession.OldCanonicalID) {
 			continue
 		}
 		kept = append(kept, succession)
@@ -1772,26 +1822,15 @@ func sourceVerdictsRecorded(sightings map[DataSource]SourceStatus) bool {
 }
 
 func (rr *ResourceRegistry) markStaleLocked(now time.Time, thresholds map[DataSource]time.Duration) {
-	thresholds = effectiveStaleThresholds(thresholds)
-
 	changed := false
 	for _, resource := range rr.resources {
 		previousStatus := resource.Status
 		staleFound := false
 		for source, status := range resource.SourceStatus {
-			threshold, ok := thresholds[source]
-			if !ok {
-				threshold = 120 * time.Second
-			}
-			if status.ExpectedUpdateIntervalSeconds > 0 {
-				// Slow inventory polls have their own cadence. A source must miss
-				// two expected intervals before its retained observation is stale.
-				threshold = max(threshold, 2*time.Duration(status.ExpectedUpdateIntervalSeconds)*time.Second)
-			}
 			if status.LastSeen.IsZero() {
 				continue
 			}
-			if now.Sub(status.LastSeen) > threshold {
+			if now.Sub(status.LastSeen) > sourceStaleThreshold(source, status, thresholds) {
 				changed = changed || status.Status != "stale"
 				status.Status = "stale"
 				resource.SourceStatus[source] = status
@@ -3940,7 +3979,7 @@ func (rr *ResourceRegistry) mergeInto(existing *Resource, incoming Resource, sou
 	existing.ParentID = rr.resolveCanonicalParentID(existing)
 
 	existing.Status = chooseStatus(existing.Status, incoming.Status, source, existing.Sources)
-	existing.Metrics = mergeMetrics(existing, existing.Metrics, incoming.Metrics, source, now, existing.SourceStatus, nil)
+	existing.Metrics = mergeMetrics(existing, existing.Metrics, incoming.Metrics, source, now, existing.SourceStatus, rr.ingestStaleThresholds)
 	existing.Metrics = clearUnavailableSourceMemoryMetric(existing.Metrics, &incoming, source)
 
 	// Prefer agent naming when available
@@ -4811,6 +4850,10 @@ func (rr *ResourceRegistry) applyManualLinks(thresholds map[DataSource]time.Dura
 
 		rr.mergeResourceData(primary, other, thresholds)
 		delete(rr.resources, otherID)
+		if rr.linkMergedIDs == nil {
+			rr.linkMergedIDs = make(map[string]struct{})
+		}
+		rr.linkMergedIDs[otherID] = struct{}{}
 		rr.updateSourceMappings(otherID, primaryID)
 	}
 }
@@ -4876,6 +4919,12 @@ func (rr *ResourceRegistry) mergeResourceData(primary *Resource, other *Resource
 	if primary.Ceph == nil {
 		primary.Ceph = other.Ceph
 	}
+	if primary.TrueNAS == nil {
+		primary.TrueNAS = other.TrueNAS
+	}
+	if primary.VMware == nil {
+		primary.VMware = other.VMware
+	}
 
 	// Manual links combine already-normalized resources. Preserve each metric's
 	// recorded source instead of flattening the linked resource to SourceAgent.
@@ -6284,6 +6333,27 @@ func mergeMetrics(
 	return &merged
 }
 
+// sourceStaleThreshold is how long a source may go without delivering before
+// its sighting is stale: the caller's threshold for the source, else the
+// default. The stale pass and the metric merge's freshness gate both read it,
+// so the gate never calls a source current that the pass marks stale, or the
+// reverse.
+func sourceStaleThreshold(source DataSource, sighting SourceStatus, thresholds map[DataSource]time.Duration) time.Duration {
+	threshold := thresholds[source]
+	if threshold <= 0 {
+		threshold = defaultStaleThresholds[source]
+	}
+	if threshold <= 0 {
+		threshold = 120 * time.Second
+	}
+	if sighting.ExpectedUpdateIntervalSeconds > 0 {
+		// Slow inventory polls have their own cadence. A source must miss
+		// two expected intervals before its retained observation is stale.
+		threshold = max(threshold, 2*time.Duration(sighting.ExpectedUpdateIntervalSeconds)*time.Second)
+	}
+	return threshold
+}
+
 // metricSourceStale reports whether a source's most recent report is older than
 // its stale threshold. A zero/unknown last-seen is treated as NOT stale so the
 // merge never demotes a source on missing information.
@@ -6300,17 +6370,7 @@ func metricSourceStale(
 	if !ok || st.LastSeen.IsZero() {
 		return false
 	}
-	threshold := time.Duration(0)
-	if configured := thresholds[source]; configured > 0 {
-		threshold = configured
-	}
-	if threshold <= 0 {
-		threshold = defaultStaleThresholds[source]
-	}
-	if threshold <= 0 {
-		threshold = 60 * time.Second
-	}
-	return now.Sub(st.LastSeen) > threshold
+	return now.Sub(st.LastSeen) > sourceStaleThreshold(source, st, thresholds)
 }
 
 func mergeMetric(

internal/unifiedresources/registry_merge_policy_test.go

@@ -247,6 +247,114 @@ func TestLeaseExpiredAgentOnLiveProxmoxNodeStaysOnline(t *testing.T) {
 	}
 }
 
+// A source merge into an existing row judges metric freshness by the ingest's
+// stale thresholds, as the stale pass that ends the ingest does. With PVE
+// polled every two minutes the monitor allows Proxmox four minutes, so a node
+// polled ninety seconds ago is current while its agent, silent for seventy-five
+// seconds, is stale, and the row shows the node's readings. Without thresholds
+// both sources are stale by the defaults and the agent keeps its precedence.
+func TestSourceMergesJudgeFreshnessByTheIngestStaleThresholds(t *testing.T) {
+	now := time.Now().UTC()
+	snapshot := models.StateSnapshot{
+		Nodes: []models.Node{{
+			ID:            "homelab-pve1",
+			Name:          "pve1",
+			Instance:      "homelab",
+			Status:        "online",
+			LinkedAgentID: "host-pve1",
+			CPU:           0.12,
+			Memory:        models.Memory{Total: 64 << 30, Used: 16 << 30, Free: 48 << 30, Usage: 25},
+			LastSeen:      now.Add(-90 * time.Second),
+		}},
+		Hosts: []models.Host{{
+			ID:              "host-pve1",
+			MachineID:       "machine-pve1",
+			Hostname:        "pve1",
+			LinkedNodeID:    "homelab-pve1",
+			Status:          "online",
+			CPUUsage:        80,
+			Memory:          models.Memory{Total: 64 << 30, Used: 48 << 30, Free: 16 << 30, Usage: 75},
+			LastSeen:        now.Add(-75 * time.Second),
+			IntervalSeconds: 30,
+		}},
+	}
+
+	for _, tc := range []struct {
+		name        string
+		thresholds  map[DataSource]time.Duration
+		proxmoxPoll string
+		want        DataSource
+		wantCPU     float64
+		wantMemory  float64
+	}{
+		{"configured", map[DataSource]time.Duration{SourceProxmox: 4 * time.Minute}, "online", SourceProxmox, 12, 25},
+		{"defaults", nil, "stale", SourceAgent, 80, 75},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			rr := NewRegistry(nil)
+			rr.IngestSnapshotWithStaleThresholds(snapshot, tc.thresholds)
+
+			resource := onlyResourceOfType(t, rr, ResourceTypeAgent)
+			if got := resource.SourceStatus[SourceProxmox].Status; got != tc.proxmoxPoll {
+				t.Fatalf("Proxmox sighting = %q, want %q: %+v", got, tc.proxmoxPoll, resource.SourceStatus)
+			}
+			if got := resource.SourceStatus[SourceAgent].Status; got != "stale" {
+				t.Fatalf("agent sighting = %q, want stale", got)
+			}
+			if resource.Metrics == nil || resource.Metrics.CPU == nil || resource.Metrics.Memory == nil {
+				t.Fatalf("merged row lost its readings: %+v", resource.Metrics)
+			}
+			if cpu := resource.Metrics.CPU; cpu.Source != tc.want || cpu.Percent != tc.wantCPU {
+				t.Fatalf("CPU = %.0f%% from %s, want %.0f%% from %s", cpu.Percent, cpu.Source, tc.wantCPU, tc.want)
+			}
+			if memory := resource.Metrics.Memory; memory.Source != tc.want || memory.Percent != tc.wantMemory {
+				t.Fatalf("memory = %.0f%% from %s, want %.0f%% from %s", memory.Percent, memory.Source, tc.wantMemory, tc.want)
+			}
+		})
+	}
+}
+
+// The metric merge and the stale pass read one threshold rule, so a source
+// sighting the merge treats as current is never one the pass marks stale.
+func TestMetricMergeFreshnessMatchesTheStalePass(t *testing.T) {
+	now := time.Now().UTC()
+	configured := map[DataSource]time.Duration{SourceProxmox: 4 * time.Minute, SourceVMware: 5 * time.Minute}
+	for _, tc := range []struct {
+		name       string
+		source     DataSource
+		sighting   SourceStatus
+		thresholds map[DataSource]time.Duration
+		wantStale  bool
+	}{
+		{"proxmox by default", SourceProxmox, SourceStatus{LastSeen: now.Add(-90 * time.Second)}, nil, true},
+		{"proxmox by configuration", SourceProxmox, SourceStatus{LastSeen: now.Add(-90 * time.Second)}, configured, false},
+		{"vsphere by default", SourceVMware, SourceStatus{LastSeen: now.Add(-3 * time.Minute)}, nil, true},
+		{"vsphere by configuration", SourceVMware, SourceStatus{LastSeen: now.Add(-3 * time.Minute)}, configured, false},
+		{"agent without its own threshold", SourceAgent, SourceStatus{LastSeen: now.Add(-75 * time.Second)}, configured, true},
+		{"slow disk inventory", SourceProxmox, SourceStatus{LastSeen: now.Add(-5 * time.Minute), ExpectedUpdateIntervalSeconds: 600}, nil, false},
+		{"missed two disk inventory polls", SourceProxmox, SourceStatus{LastSeen: now.Add(-21 * time.Minute), ExpectedUpdateIntervalSeconds: 600}, nil, true},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			status := map[DataSource]SourceStatus{tc.source: tc.sighting}
+			if got := metricSourceStale(now, status, tc.source, tc.thresholds); got != tc.wantStale {
+				t.Fatalf("metric merge stale = %v, want %v", got, tc.wantStale)
+			}
+
+			rr := NewRegistry(nil)
+			rr.resources["resource"] = &Resource{
+				ID:           "resource",
+				Status:       StatusOnline,
+				Sources:      []DataSource{tc.source},
+				SourceStatus: map[DataSource]SourceStatus{tc.source: tc.sighting},
+			}
+			rr.MarkStale(now, tc.thresholds)
+			if got := rr.resources["resource"].SourceStatus[tc.source].Status == "stale"; got != tc.wantStale {
+				t.Fatalf("stale pass stale = %v, want %v", got, tc.wantStale)
+			}
+		})
+	}
+}
+
 // Docker hosts hold a shorter lease than the registry's Docker stale
 // threshold, so a silent Docker host was offline for a while and then
 // flipped to warning once its sighting went stale.
@@ -441,6 +549,319 @@ func TestManualLinkKeepsTheLiveSightingOfASharedSource(t *testing.T) {
 	}
 }
 
+// Either side of an operator link can arrive as a supplemental record: a
+// vSphere VM with a Pulse agent inside, or an agent on a TrueNAS host. The
+// monitor's rebuild ingests records after the snapshot, so it must join them
+// there too, keep the record side's provider payload, and list what the
+// resources API lists after seeding from it.
+func TestManualLinksJoinRecordIngestedResourcesInMonitorRebuild(t *testing.T) {
+	now := time.Now().UTC()
+	snapshot := models.StateSnapshot{
+		Hosts: []models.Host{
+			{ID: "host-app-guest", Hostname: "app-guest", MachineID: "machine-app-guest", Status: "online", LastSeen: now},
+			{ID: "host-nas", Hostname: "nas-a-mgmt", MachineID: "machine-nas-a", Status: "online", LastSeen: now},
+		},
+		LastUpdate: now,
+	}
+	vmRecord := IngestRecord{
+		SourceID: "vc-1:vm:vm-42",
+		Resource: Resource{
+			Type:       ResourceTypeVM,
+			Technology: "vmware",
+			Name:       "app-guest",
+			Status:     StatusOnline,
+			LastSeen:   now,
+			VMware:     &VMwareData{ConnectionID: "vc-1", ManagedObjectID: "vm-42", EntityType: "vm"},
+		},
+		Identity: ResourceIdentity{Hostnames: []string{"app-guest"}},
+	}
+	records := map[DataSource][]IngestRecord{
+		SourceVMware: {vmRecord},
+		SourceTrueNAS: {
+			{
+				SourceID: "system:tn-1",
+				Resource: Resource{
+					Type:     ResourceTypeAgent,
+					Name:     "nas-a",
+					Status:   StatusOnline,
+					LastSeen: now,
+					TrueNAS:  &TrueNASData{Hostname: "nas-a"},
+				},
+				Identity: ResourceIdentity{Hostnames: []string{"nas-a"}},
+			},
+			{
+				SourceID:       "system:tn-1:pool:tank",
+				ParentSourceID: "system:tn-1",
+				Resource: Resource{
+					Type:     ResourceTypeStorage,
+					Name:     "tank",
+					Status:   StatusOnline,
+					LastSeen: now,
+					Storage:  &StorageMeta{Type: "zfs-pool", Platform: "truenas", Topology: "pool"},
+				},
+			},
+		},
+	}
+
+	for _, agentPrimary := range []bool{false, true} {
+		name := "record-primary"
+		if agentPrimary {
+			name = "agent-primary"
+		}
+		t.Run(name, func(t *testing.T) {
+			store := NewMemoryStore()
+			adapter := NewMonitorAdapter(NewRegistry(store))
+			adapter.PopulateSnapshotAndSupplemental(snapshot, records)
+
+			var vmID, guestAgentID, systemID, nasAgentID string
+			for _, resource := range adapter.GetAll() {
+				switch {
+				case resource.VMware != nil:
+					vmID = resource.ID
+				case resource.TrueNAS != nil:
+					systemID = resource.ID
+				case resource.Name == "app-guest":
+					guestAgentID = resource.ID
+				case resource.Name == "nas-a-mgmt":
+					nasAgentID = resource.ID
+				}
+			}
+			if vmID == "" || guestAgentID == "" || systemID == "" || nasAgentID == "" {
+				t.Fatalf("unlinked estate = %v, want the VM, both agents and the TrueNAS system", resourceIDs(adapter.GetAll()))
+			}
+			guestPrimary, nasPrimary := vmID, systemID
+			if agentPrimary {
+				guestPrimary, nasPrimary = guestAgentID, nasAgentID
+			}
+			for _, link := range []ResourceLink{
+				{ResourceA: vmID, ResourceB: guestAgentID, PrimaryID: guestPrimary},
+				{ResourceA: systemID, ResourceB: nasAgentID, PrimaryID: nasPrimary},
+			} {
+				if err := store.AddLink(link); err != nil {
+					t.Fatalf("add link: %v", err)
+				}
+			}
+
+			assertJoined := func(stage string) {
+				t.Helper()
+				resources := adapter.GetAll()
+				if len(resources) != 3 {
+					t.Fatalf("%s: resources = %v, want the linked VM, the linked TrueNAS system and its pool", stage, resourceIDs(resources))
+				}
+				// An agent inside a guest supplements the guest, whichever
+				// side the operator chose as primary.
+				guest, ok := adapter.currentRegistry().Get(vmID)
+				if !ok {
+					t.Fatalf("%s: linked vSphere VM %s missing from %v", stage, vmID, resourceIDs(resources))
+				}
+				if guest.Type != ResourceTypeVM || guest.VMware == nil || guest.Agent == nil {
+					t.Fatalf("%s: linked VM type=%s vmware=%t agent=%t, want the VM carrying its agent", stage, guest.Type, guest.VMware != nil, guest.Agent != nil)
+				}
+				if !slices.Contains(guest.Sources, SourceVMware) || !slices.Contains(guest.Sources, SourceAgent) {
+					t.Fatalf("%s: linked VM sources = %v, want vmware and agent", stage, guest.Sources)
+				}
+				system, ok := adapter.currentRegistry().Get(nasPrimary)
+				if !ok {
+					t.Fatalf("%s: linked TrueNAS system %s missing from %v", stage, nasPrimary, resourceIDs(resources))
+				}
+				if system.TrueNAS == nil || system.Agent == nil {
+					t.Fatalf("%s: linked system truenas=%t agent=%t, want the primary carrying both payloads", stage, system.TrueNAS != nil, system.Agent != nil)
+				}
+				if !slices.Contains(system.Sources, SourceTrueNAS) || !slices.Contains(system.Sources, SourceAgent) {
+					t.Fatalf("%s: linked system sources = %v, want truenas and agent", stage, system.Sources)
+				}
+				pool := onlyResourceOfType(t, adapter.currentRegistry(), ResourceTypeStorage)
+				if pool.ParentID == nil || *pool.ParentID != nasPrimary {
+					t.Fatalf("%s: pool parent = %v, want the linked system %s", stage, pool.ParentID, nasPrimary)
+				}
+
+				// Identity pins persisted by the rebuild must not re-key a link
+				// onto its own primary.
+				links, err := store.GetLinks()
+				if err != nil {
+					t.Fatalf("%s: get links: %v", stage, err)
+				}
+				for _, link := range links {
+					if link.ResourceA == link.ResourceB {
+						t.Fatalf("%s: link collapsed onto %s", stage, link.ResourceA)
+					}
+				}
+
+				// The resources API seeds its registry from this listing
+				// through resource ingest and must agree with it.
+				rest := NewRegistry(store)
+				rest.IngestResources(resources)
+				if got, want := resourceIDs(rest.List()), resourceIDs(resources); !slices.Equal(got, want) {
+					t.Fatalf("%s: a registry seeded from the monitor lists %v, monitor lists %v", stage, got, want)
+				}
+			}
+
+			next := snapshot
+			next.LastUpdate = now.Add(time.Second)
+			adapter.PopulateSnapshotAndSupplemental(next, records)
+			assertJoined("rebuild")
+
+			adapter.PopulateSupplementalRecords(SourceVMware, records[SourceVMware])
+			adapter.PopulateSupplementalRecords(SourceTrueNAS, records[SourceTrueNAS])
+			assertJoined("supplemental refresh")
+
+			next.LastUpdate = now.Add(2 * time.Second)
+			adapter.PopulateSnapshotAndSupplemental(next, records)
+			assertJoined("second rebuild")
+		})
+	}
+}
+
+// A Proxmox node that absorbs a linked agent carries the agent's machine key,
+// so the rebuild pins the node with it. The agent is still observed: that pin
+// must not succeed the agent's ID onto the node, which re-keyed the link onto
+// the node itself and split the pair on the next rebuild.
+func TestManualLinkKeepsItsEndpointsThroughIdentityPinPersistence(t *testing.T) {
+	now := time.Now().UTC()
+	snapshot := models.StateSnapshot{
+		Nodes: []models.Node{{
+			ID: "lab-pve1", Name: "pve1", Instance: "lab",
+			Host: "https://192.0.2.10:8006", Status: "online", LastSeen: now,
+		}},
+		Hosts: []models.Host{{
+			ID: "host-box", Hostname: "box-x", MachineID: "machine-box", Status: "online", LastSeen: now,
+		}},
+		LastUpdate: now,
+	}
+	store := NewMemoryStore()
+	adapter := NewMonitorAdapter(NewRegistry(store))
+	adapter.PopulateSnapshotAndSupplemental(snapshot, nil)
+	var nodeID, agentID string
+	for _, resource := range adapter.GetAll() {
+		if resource.Proxmox != nil {
+			nodeID = resource.ID
+		} else if resource.Agent != nil {
+			agentID = resource.ID
+		}
+	}
+	if nodeID == "" || agentID == "" || nodeID == agentID {
+		t.Fatalf("unlinked estate = %v, want a node and a separate agent", resourceIDs(adapter.GetAll()))
+	}
+	if err := store.AddLink(ResourceLink{ResourceA: nodeID, ResourceB: agentID, PrimaryID: nodeID}); err != nil {
+		t.Fatalf("add link: %v", err)
+	}
+
+	for rebuild := 1; rebuild <= 3; rebuild++ {
+		next := snapshot
+		next.LastUpdate = now.Add(time.Duration(rebuild) * time.Second)
+		adapter.PopulateSnapshotAndSupplemental(next, nil)
+		if got := resourceIDs(adapter.GetAll()); !slices.Equal(got, []string{nodeID}) {
+			t.Fatalf("rebuild %d: resources = %v, want only the linked node %s", rebuild, got, nodeID)
+		}
+		links, err := store.GetLinks()
+		if err != nil {
+			t.Fatalf("get links: %v", err)
+		}
+		if len(links) != 1 || links[0].ResourceA != nodeID || links[0].ResourceB != agentID {
+			t.Fatalf("rebuild %d: links = %+v, want %s -> %s kept", rebuild, links, nodeID, agentID)
+		}
+	}
+}
+
+// A record may name an older canonical ID as superseded on every ingest. When
+// that ID belongs to a resource a manual link folded into the record's own
+// resource, it is still observed: succeeding it on a later refresh would
+// re-key the link onto its primary.
+func TestRecordDeclaredSuccessionSparesALinkFoldedResource(t *testing.T) {
+	now := time.Now().UTC()
+	snapshot := models.StateSnapshot{
+		Hosts:      []models.Host{{ID: "host-nas", Hostname: "nas-a-mgmt", MachineID: "machine-nas-a", Status: "online", LastSeen: now}},
+		LastUpdate: now,
+	}
+	agentID := MachineIdentityCanonicalID(ResourceTypeAgent, "machine-nas-a")
+	records := []IngestRecord{{
+		SourceID:               "system:tn-1",
+		SupersededCanonicalIDs: []string{agentID},
+		Resource: Resource{
+			Type:     ResourceTypeAgent,
+			Name:     "nas-a",
+			Status:   StatusOnline,
+			LastSeen: now,
+			TrueNAS:  &TrueNASData{Hostname: "nas-a"},
+		},
+		Identity: ResourceIdentity{Hostnames: []string{"nas-a"}},
+	}}
+
+	unlinked := NewMonitorAdapter(NewRegistry(nil))
+	unlinked.PopulateSnapshotAndSupplemental(snapshot, map[DataSource][]IngestRecord{SourceTrueNAS: records})
+	var systemID string
+	for _, resource := range unlinked.GetAll() {
+		if resource.TrueNAS != nil {
+			systemID = resource.ID
+		}
+	}
+	if systemID == "" || !slices.Contains(resourceIDs(unlinked.GetAll()), agentID) {
+		t.Fatalf("unlinked estate = %v, want agent %s and a TrueNAS system", resourceIDs(unlinked.GetAll()), agentID)
+	}
+
+	store := NewMemoryStore()
+	if err := store.AddLink(ResourceLink{ResourceA: systemID, ResourceB: agentID, PrimaryID: systemID}); err != nil {
+		t.Fatalf("add link: %v", err)
+	}
+	adapter := NewMonitorAdapter(NewRegistry(store))
+	adapter.PopulateSnapshotAndSupplemental(snapshot, map[DataSource][]IngestRecord{SourceTrueNAS: records})
+	adapter.PopulateSupplementalRecords(SourceTrueNAS, records)
+
+	if got := resourceIDs(adapter.GetAll()); !slices.Equal(got, []string{systemID}) {
+		t.Fatalf("resources = %v, want only the linked system %s", got, systemID)
+	}
+	links, err := store.GetLinks()
+	if err != nil {
+		t.Fatalf("get links: %v", err)
+	}
+	if len(links) != 1 || links[0].ResourceA != systemID || links[0].ResourceB != agentID {
+		t.Fatalf("links = %+v, want %s -> %s kept", links, systemID, agentID)
+	}
+}
+
+// A link merge folds the other resource's provider payloads into the primary
+// where the primary has none, whichever side the operator chose.
+func TestManualLinkKeepsProviderPayloadsThePrimaryLacks(t *testing.T) {
+	now := time.Now().UTC()
+	store := NewMemoryStore()
+	for _, link := range []ResourceLink{
+		{ResourceA: "agent-nas", ResourceB: "truenas-system", PrimaryID: "agent-nas"},
+		{ResourceA: "agent-esxi", ResourceB: "vmware-host", PrimaryID: "agent-esxi"},
+	} {
+		if err := store.AddLink(link); err != nil {
+			t.Fatalf("add link: %v", err)
+		}
+	}
+	host := func(id, name string, sources ...DataSource) Resource {
+		status := make(map[DataSource]SourceStatus, len(sources))
+		for _, source := range sources {
+			status[source] = SourceStatus{Status: "online", LastSeen: now}
+		}
+		return Resource{ID: id, Type: ResourceTypeAgent, Name: name, Status: StatusOnline, LastSeen: now, Sources: sources, SourceStatus: status}
+	}
+	agentNAS := host("agent-nas", "nas-a-mgmt", SourceAgent)
+	agentNAS.Agent = &AgentData{AgentID: "agent-nas", Hostname: "nas-a-mgmt"}
+	system := host("truenas-system", "nas-a", SourceTrueNAS)
+	system.TrueNAS = &TrueNASData{Hostname: "nas-a"}
+	agentESXi := host("agent-esxi", "esxi-01", SourceAgent)
+	agentESXi.Agent = &AgentData{AgentID: "agent-esxi", Hostname: "esxi-01"}
+	esxi := host("vmware-host", "esxi-01.lab", SourceVMware)
+	esxi.VMware = &VMwareData{ConnectionID: "vc-1", ManagedObjectID: "host-101", EntityType: "host"}
+
+	rr := NewRegistry(store)
+	rr.IngestResources([]Resource{agentNAS, system, agentESXi, esxi})
+
+	if got := resourceIDs(rr.List()); !slices.Equal(got, []string{"agent-esxi", "agent-nas"}) {
+		t.Fatalf("resources = %v, want the two linked primaries", got)
+	}
+	if nas, _ := rr.Get("agent-nas"); nas.TrueNAS == nil || nas.Agent == nil {
+		t.Fatalf("linked TrueNAS system truenas=%t agent=%t, want both payloads", nas.TrueNAS != nil, nas.Agent != nil)
+	}
+	if host, _ := rr.Get("agent-esxi"); host.VMware == nil || host.Agent == nil {
+		t.Fatalf("linked vSphere host vmware=%t agent=%t, want both payloads", host.VMware != nil, host.Agent != nil)
+	}
+}
+
 func TestLeaseExpiredKubernetesClusterStaysOffline(t *testing.T) {
 	rr := NewRegistry(nil)
 	rr.IngestSnapshot(models.StateSnapshot{