hub v0.110.0: allowlist + operator-only for offsite_proof_empty (R-87)

Controller v0.231.0 adds a nightly job that proves an app's newest off-site snapshot still
CONTAINS that app's data. When it finds one that does not, it emits offsite_proof_empty.

Two register lines, both load-bearing and both in this commit:
- allowedEventTypes - an unallowlisted type is answered 400 and VANISHES, so this entry is
  what makes the alarm exist at all.
- operatorOnlyEvents - a missing customerMessages entry is NOT a routing block
  (FormatCustomerEmail falls back to the raw message, the v0.78.0 defect that register was
  built for). A customer can take no action on a hollow recovery unit.

DELIBERATELY NOT reusing backup_integrity_failed, which is the nearest existing type: it
means THE STORE IS DAMAGED and carries the Hungarian template saying so. Here the store is
sound and the CONTENT is absent - different cause, different action, and telling a customer
their backups are damaged when they are not is the more expensive mistake. Same asymmetry
looksLikeRepositoryDamage is shaped around.

DELIBERATELY no customerMessages entry (the controller's dynamic Hungarian names the app and
what is missing; a template would discard it) and DELIBERATELY not in perAppCooldownEvents
(a fenced act - this job proves ONE app per night, so the coarse hourly cooldown is already
the right grain).

This widens the R-87 task's stated repo scope to felhom.eu/hub/. The reason is recorded in
felhom-controller/CONTEXT.md ruling 4 rather than left as an unexplained diff.

Hub green gate: go build/vet/test all pass, 18 packages.

hub/CHANGELOG.md

@@ -1,3 +1,32 @@
+## v0.110.0 — `offsite_proof_empty`: a backup that is intact and holds nothing is not a damaged store (2026-08-31, R-87)
+
+**Two lines of register, and the reason they are not one line of reuse.** Controller v0.231.0 adds a
+nightly job that proves an app's newest off-site snapshot still CONTAINS that app's data. When it
+finds one that does not, it needs to say so — and the nearest existing type, `backup_integrity_failed`,
+would say the wrong thing.
+
+- **`allowedEventTypes` gains `offsite_proof_empty`.** An event type the hub does not allowlist is
+  answered **400 and vanishes** (R-97a / the event-allowlist gotcha), so this entry is what makes the
+  alarm exist at all.
+- **`operatorOnlyEvents` gains it in the SAME commit.** A missing `customerMessages` entry is NOT a
+  routing block — `FormatCustomerEmail` falls back to the raw message, which is the v0.78.0 defect
+  that register was built for. The customer can take no action on a hollow recovery unit: the causes
+  are capture-side and diagnosed from the manifest and the capture legs, both operator surfaces. The
+  customer's actionable half of this class is the fill warning, which fires first.
+- **Deliberately NO `customerMessages` entry.** The controller sends a dynamic Hungarian message
+  naming the app and what is missing; a template would discard those specifics — the
+  `offbox_enlarge_blocked` / `disk_health_degraded` precedent.
+- **Deliberately NOT in `perAppCooldownEvents`.** That is a fenced act (08 §6.2): the backup family's
+  cooldown is coarse on purpose. This job proves ONE app per night, so the hourly operator cooldown is
+  already the right grain and an entry would widen it with no cause.
+
+**Why not reuse `backup_integrity_failed`:** it means *the store is damaged* and carries the Hungarian
+template that says so. Here the store is sound and the CONTENT is absent — a different cause, a
+different action, and telling a customer their backups are damaged when they are not is the more
+expensive mistake. That asymmetry is the same one `looksLikeRepositoryDamage` is shaped around.
+
+---
+
 ## v0.109.0 — the Backup card told every operator that every customer had no backups (2026-08-30, R-331)
 
 > **This push used `git push --no-verify`, and that is declared here rather than worked around.**

hub/internal/api/handler.go

@@ -2005,6 +2005,20 @@ var allowedEventTypes = map[string]bool{
 	"db_dump_failed":          true,
 	"backup_integrity_ok":     true,
 	"backup_integrity_failed": true,
+	// controller v0.231.0 (R-87) — the nightly off-site PROOF found a backup that is READABLE and
+	// contains none of the app's data. DELIBERATELY NOT `backup_integrity_failed`, which is the
+	// nearest existing type and would be the wrong sentence: that one means the STORE IS DAMAGED and
+	// carries a Hungarian template saying the integrity check found an error. Here the store is sound
+	// and the CONTENT is missing — a different fact, a different cause and a different action.
+	// Telling a customer their backups are damaged when they are not is the more expensive mistake,
+	// and it is the one R-359's own `looksLikeRepositoryDamage` is shaped to avoid.
+	//
+	// NO `customerMessages` entry, deliberately: the controller sends a dynamic Hungarian message
+	// naming the app and what is missing, and a template would DISCARD those specifics
+	// (templates.go priority) — the `offbox_enlarge_blocked` / `disk_health_degraded` precedent.
+	// Operator routing is enforced by `notify.operatorOnlyEvents`, in this same commit; this entry
+	// alone does NOT make it operator-only (the v0.78.0 defect that register records).
+	"offsite_proof_empty": true,
 	"crossdrive_completed":    true,
 	"crossdrive_failed":       true,
 	// controller v0.134.1 — enlarged offsite push refused by the quota gate (warning; the controller's

hub/internal/notify/dispatcher.go

@@ -517,6 +517,16 @@ var operatorOnlyEvents = map[string]bool{
 	// block — `FormatCustomerEmail` falls back to the raw English message. That mistake shipped
 	// once (v0.78.0) and the comment above records it.
 	"backup_run_failures": true,
+	// R-87 (controller v0.231.0). The nightly off-site proof found a backup that is intact and holds
+	// none of the app's data. Operator-only for the same reason as `recovery_unit_capture_failed`
+	// above: the customer can take no action on it — a hollow recovery unit is a product/ops fault,
+	// diagnosed from the capture legs and the manifest, both of which are operator surfaces. The
+	// customer's actionable half of this class is the fill warning, which fires first.
+	//
+	// LISTED IN THE SAME COMMIT THAT MINTS THE TYPE, because a missing customerMessages entry is NOT
+	// a block — FormatCustomerEmail falls back to the raw message, which is the v0.78.0 defect this
+	// register was built for.
+	"offsite_proof_empty": true,
 	// R-197 (v0.93.0). "The sealed offsite repository key changed" is a custody fact about escrow
 	// blobs. A customer can take no action on it — the remedy is the operator's inspection of the
 	// off-site tier — and the text is operator-grade English naming host ids and retained-blob