diff --git a/.github/workflows/fetch-telemetry.yml b/.github/workflows/fetch-telemetry.yml
index fd25bf1b..fd25ef27 100644
--- a/.github/workflows/fetch-telemetry.yml
+++ b/.github/workflows/fetch-telemetry.yml
@@ -1,9 +1,13 @@
 name: Fetch Telemetry Data
 
 on:
-  # Keep this as a manual telemetry-only backfill path. The monthly OSS Health
-  # refresh intentionally does not update telemetry while /api/overview differs
-  # from the Grafana-backed source of truth.
+  # Monthly telemetry refresh. /api/overview now counts entities over the period
+  # window instead of at an instant (cozystack/cozystack-telemetry-server#8), so
+  # it matches the Grafana-backed source of truth and is safe to consume on a
+  # schedule again. Aligned with the monthly OSS Health cron; workflow_dispatch
+  # remains for manual backfills.
+  schedule:
+    - cron: '0 4 1 * *'
   workflow_dispatch:
 
 permissions:
diff --git a/hack/fetch_telemetry.py b/hack/fetch_telemetry.py
index 2f9d7d8e..06f0a64c 100755
--- a/hack/fetch_telemetry.py
+++ b/hack/fetch_telemetry.py
@@ -8,8 +8,9 @@
 3. Merge case-insensitive / Pax* / legacy-name aliases into one canonical entry
    per application, keeping the maximum instance count (zero-count entries
    left after the merge are dropped from the table).
-4. Pull `Tenant` out of the apps map and surface it as the top-level Tenants
-   summary card (the raw `total_tenants` field from the API is always zero).
+4. Surface tenant count as the top-level Tenants summary card, preferring the
+   period's `total_tenants` field (correct since telemetry-server #8) and
+   falling back to `apps.Tenant` for older payloads where it was zero.
 5. Emit the payload in the shape consumed by `oss-health-app.html` +
    `renderTelemetry`, including `summary_cards`, `apps`, `range`.
 
@@ -100,7 +101,9 @@ def transform_period(raw_period: dict, label_fallback: str) -> dict | None:
     if not raw_period:
         return None
     apps_raw = raw_period.get("apps", {}) or {}
-    tenants = int(apps_raw.get("Tenant", 0))
+    # Prefer the period's own total_tenants (correct since telemetry-server #8);
+    # fall back to apps.Tenant for older payloads where total_tenants was 0.
+    tenants = int(raw_period.get("total_tenants") or apps_raw.get("Tenant", 0))
     clusters = int(raw_period.get("clusters", 0))
     total_nodes = int(raw_period.get("total_nodes", 0))
     avg_nodes = raw_period.get("avg_nodes_per_cluster")