diff --git a/.github/workflows/pr-review-merge-scheduler.yml b/.github/workflows/pr-review-merge-scheduler.yml
index b3deb32eef..a15cdf36e1 100644
--- a/.github/workflows/pr-review-merge-scheduler.yml
+++ b/.github/workflows/pr-review-merge-scheduler.yml
@@ -74,15 +74,15 @@ on:
         type: string
   schedule:
     - cron: "*/30 * * * *"
-    # Every-15-minutes org-wide sweep cadence for the org-queue-sweep job below. Target
+    # Hourly org-wide sweep cadence for the org-queue-sweep job below. Target
     # repositories only receive scheduler runs on PR events, review/security
     # workflow completion, and protected-branch pushes; a PR whose approval or
     # required checks land AFTER its last event has no later trigger and sits
     # approved-but-unmerged until a human pushes something. The sweep closes
-    # that gap on a fixed heartbeat. Runs every 15 minutes so an approval or
+    # that gap on a fixed heartbeat. Runs hourly so an approval or
     # required check that lands after a PR's last event is auto-updated/merged
-    # within ~15 minutes instead of sitting idle for up to an hour.
-    - cron: "*/15 * * * *"
+    # within about an hour without adding quarter-hourly runner pressure.
+    - cron: "0 * * * *"
   repository_dispatch:
     types: [merge-scheduler]
 
@@ -136,7 +136,7 @@ jobs:
       ) &&
       (
         github.event_name != 'schedule' ||
-        github.event.schedule != '*/15 * * * *'
+        github.event.schedule != '0 * * * *'
       ) &&
       (
         github.event_name != 'repository_dispatch' ||
@@ -587,12 +587,12 @@ jobs:
     if: >-
       github.repository == 'ContextualWisdomLab/.github' &&
       (
-        (github.event_name == 'schedule' && github.event.schedule == '*/15 * * * *') ||
+        (github.event_name == 'schedule' && github.event.schedule == '0 * * * *') ||
         (github.event_name == 'repository_dispatch' && github.event.client_payload.org_sweep == true)
       )
     runs-on: ubuntu-24.04
     # The complete organization walk exceeded the legacy 30-minute boundary in
-    # production. Keep one running and one latest pending */15 sweep through the
+    # production. Keep one running and one latest pending hourly sweep through the
     # schedule-specific concurrency key above, while allowing the current walk
     # enough time to finish instead of cancelling before later repositories.
     timeout-minutes: 60
@@ -636,7 +636,7 @@ jobs:
       # ticks" guarantee a rotation is meant to provide. Wall-clock time alone
       # is also insufficient, since this single-flight/non-cancelling job can
       # run up to 60 minutes and a delayed real execution can let more than
-      # one 900s window elapse, occasionally repeating a modulo offset
+      # one hourly window elapse, occasionally repeating a modulo offset
       # (ContextualWisdomLab/.github#1223 review finding).
       # A repository the sweep credential structurally cannot read (the OpenCode
       # app is not installed there / the PR_REVIEW_MERGE_TOKEN lacks it) returns
@@ -850,10 +850,10 @@ jobs:
           # source: a persistent `ORG_SWEEP_ROTATION_COUNTER` repository
           # variable on this (.github) repository, incremented by exactly
           # one at the start of every actual org-queue-sweep execution. A
-          # wall-clock tick (one per 900s) is *not* sufficient on its own:
+          # wall-clock tick (one per hour) is *not* sufficient on its own:
           # this job is single-flight/non-cancelling with up to a 60-minute
           # timeout, so a delayed or backlogged execution can let more than
-          # one 900s window elapse between two real sweep runs, and if that
+          # one hourly window elapse between two real sweep runs, and if that
           # gap happens to be an exact multiple of the repository count the
           # modulo offset repeats -- reintroducing the exact starvation
           # #1220 fixed (CodeRabbit review finding on #1223). A persistent
@@ -917,7 +917,7 @@ jobs:
                 ORG_SWEEP_ROTATION_INDEX="$counter_next"
               else
                 echo "::warning::read ${counter_variable_name}=${counter_current} but could not PATCH it; falling back to a wall-clock rotation tick for this run only"
-                ORG_SWEEP_ROTATION_INDEX=$(( $(date -u +%s) / 900 ))
+                ORG_SWEEP_ROTATION_INDEX=$(( $(date -u +%s) / 3600 ))
               fi
             elif gh api "repos/${GITHUB_REPOSITORY}/actions/variables" \
                    -X POST -f "name=${counter_variable_name}" -f "value=1" >/dev/null 2>&1; then
@@ -930,7 +930,7 @@ jobs:
               ORG_SWEEP_ROTATION_INDEX=1
             else
               echo "::warning::could not read/write ${counter_variable_name}; falling back to a wall-clock rotation tick for this run only"
-              ORG_SWEEP_ROTATION_INDEX=$(( $(date -u +%s) / 900 ))
+              ORG_SWEEP_ROTATION_INDEX=$(( $(date -u +%s) / 3600 ))
             fi
           fi
           if ! [[ "$ORG_SWEEP_ROTATION_INDEX" =~ ^[0-9]+$ ]]; then
diff --git a/docs/doctoring/actions-queue-saturation-hourly-sweep.md b/docs/doctoring/actions-queue-saturation-hourly-sweep.md
new file mode 100644
index 0000000000..a0d3122290
--- /dev/null
+++ b/docs/doctoring/actions-queue-saturation-hourly-sweep.md
@@ -0,0 +1,36 @@
+# Actions queue saturation: hourly organization sweep
+
+**Status:** active repair evidence
+**Owning repository:** `ContextualWisdomLab/.github`
+**Canonical repair PR:** `#1630`
+**Protected baseline:** `main@4ae90e18b03a3a455e13e501628010cabc5c37a8`
+
+## Root cause
+
+The central PR review/merge scheduler has two periodic entry points in addition to event-driven wakes. The repository-local queue scan runs every 30 minutes, while the expensive `org-queue-sweep` has been admitted every 15 minutes. Under the observed organization-wide hosted-runner saturation, the full organization walk can remain queued or run long enough that quarter-hourly admission adds more pending work before prior evidence drains. That is a control-plane pressure amplifier: required current-head evidence for leaf repositories queues behind recurring control-plane work that exists to unblock those same repositories.
+
+The repair is deliberately bounded. Keep the 30-minute repository scan and all event-driven `pull_request_target`, `pull_request_review`, `workflow_run`, and `repository_dispatch` wakes. Change only the organization sweep heartbeat to hourly (`0 * * * *`). The wall-clock fallback used by the persisted sweep rotation counter must advance on the same hourly cadence (`epoch_seconds / 3600`) rather than the old 15-minute cadence (`epoch_seconds / 900`), otherwise a fallback run would skip four repository offsets for each real scheduled sweep.
+
+## TDD and executable contract
+
+`tests/test_actions_queue_saturation_scheduler_cadence.py` is the RED-first contract. It requires the live workflow to contain the hourly cron, rejects the quarter-hourly cron, preserves event-driven wakes, and binds both wall-clock fallback expressions to hourly rotation. The older assertions in `tests/test_required_workflow_queue_contract.py` must be updated with the production workflow rather than retained as a stale policy test.
+
+The production change must also update `docs/org-required-workflow-rollout.md` so operator guidance states that the heartbeat can be up to one hour old. Historical doctoring that describes the old quarter-hour schedule remains historical evidence and must not be rewritten as though it never existed.
+
+## Safety boundary
+
+This repair does not mark queued checks successful, cancel the sole current-head evidence, weaken required workflows, relax approval requirements, or synthesize review state. Queue hygiene remains fail-closed. Cross-repository mutation credentials, exact-head validation, stale-head cancellation rules, unavailable-repository thresholds, scheduler concurrency groups, and merge guards remain unchanged.
+
+No organization-owned identifier introduced by this repair uses an ambiguous single-word domain name. GitHub event fields and cron syntax are externally mandated contract terms and remain unchanged except for the cadence value.
+
+## Verification
+
+After the production commit lands on the canonical branch:
+
+1. run the focused cadence and required-workflow queue contract tests;
+2. verify the scheduler workflow contains exactly the intended 30-minute repository scan and hourly organization sweep;
+3. confirm event-driven wakes remain present;
+4. inspect fresh exact-head required checks and review evidence;
+5. observe queue depth after the change rather than treating the configuration diff itself as proof that saturation has cleared.
+
+Merge remains subject to ordinary protected-branch requirements and exact-current-head evidence.
diff --git a/docs/org-required-workflow-rollout.md b/docs/org-required-workflow-rollout.md
index 36edcd29dd..7c55c6fbab 100644
--- a/docs/org-required-workflow-rollout.md
+++ b/docs/org-required-workflow-rollout.md
@@ -156,9 +156,9 @@ The central `.github/workflows/pr-review-merge-scheduler.yml` is now part of the
 
 Do not centralize the scheduler by running a `.github` scheduled job against other repositories with the `.github` repository token. That would either fail permission checks or use the wrong mutation actor. The central path is a required workflow executed in each target repository context.
 
-- Heartbeat fallback posture: event-driven target-repository runs stop retrying once their triggering event is consumed, so a PR that becomes mergeable AFTER its last event (approval published after the scheduler pass, merge-preview checks landing late, a temporary base-branch policy blocker clearing) has no later trigger and sits approved-but-unmerged. The `org-queue-sweep` job in the central scheduler workflow closes this gap: it runs every 15 minutes (`*/15 * * * *`) only in `ContextualWisdomLab/.github`, re-runs the same trusted scheduler script against every non-archived organization repository, and merges/updates through the identical guarded contract. Stacked PRs, which do not receive injected required workflows, use a separate bounded OpenCode dispatch budget so ordinary default-branch traffic cannot leave them at `OpenCode review absent`. It never uses the `.github` repository `github.token` for sibling mutations — it requires `PR_REVIEW_MERGE_TOKEN`, `OPENCODE_APPROVE_TOKEN`, or the exchanged OpenCode app token, and fails with a visible `::error` reason when no cross-repository mutation credential is available instead of silently no-opping. Every swept repository prints its per-PR decision log, so an unmerged PR always has a concrete logged reason at most 15 minutes old.
+- Heartbeat fallback posture: event-driven target-repository runs stop retrying once their triggering event is consumed, so a PR that becomes mergeable AFTER its last event (approval published after the scheduler pass, merge-preview checks landing late, a temporary base-branch policy blocker clearing) has no later trigger and sits approved-but-unmerged. The `org-queue-sweep` job in the central scheduler workflow closes this gap: it runs hourly (`0 * * * *`) only in `ContextualWisdomLab/.github`, re-runs the same trusted scheduler script against every non-archived organization repository, and merges/updates through the identical guarded contract. Stacked PRs, which do not receive injected required workflows, use a separate bounded OpenCode dispatch budget so ordinary default-branch traffic cannot leave them at `OpenCode review absent`. It never uses the `.github` repository `github.token` for sibling mutations — it requires `PR_REVIEW_MERGE_TOKEN`, `OPENCODE_APPROVE_TOKEN`, or the exchanged OpenCode app token, and fails with a visible `::error` reason when no cross-repository mutation credential is available instead of silently no-opping. Every swept repository prints its per-PR decision log, so an unmerged PR always has a concrete logged reason at most one hour old.
 - Queue hygiene posture: during the sweep, workflow runs still `queued` after `ORG_SWEEP_STALE_QUEUE_HOURS` (default 24h) are cancelled with their run id, workflow name, head branch, and age logged. A run queued that long belongs to a head that PR events will never revisit (closed PR, force-pushed branch, or a previous runner outage), and leaving it keeps the Actions queue holding non-current-head work.
-- Inaccessible-repository posture: a sibling repository the sweep credential structurally cannot read — the OpenCode app is not installed there, or `PR_REVIEW_MERGE_TOKEN` does not cover it — returns HTTP 403 `Resource not accessible by integration` on every read. That is an access-grant fact the automation can never resolve, so the sweep classifies it as a skipped, non-fatal **unavailable** repository (a `::warning` naming the repository and the remediation) instead of a hard failure. Without this, a handful of un-enrolled repositories keeps the scheduled sweep heartbeat (the org sweep's `*/15 * * * *` cron) permanently red and masks a genuinely new repository that starts failing. Fail-closed is preserved on both sides: any non-403 scheduler failure still fails the sweep with its per-PR reason, and if more than `ORG_SWEEP_MAX_UNAVAILABLE` (default 5) repositories become unreachable in one pass — a credential-scope regression rather than a few un-enrolled repos — the job fails loudly. Remediation for a listed repository is to install the OpenCode app on it or grant `PR_REVIEW_MERGE_TOKEN` access.
+- Inaccessible-repository posture: a sibling repository the sweep credential structurally cannot read — the OpenCode app is not installed there, or `PR_REVIEW_MERGE_TOKEN` does not cover it — returns HTTP 403 `Resource not accessible by integration` on every read. That is an access-grant fact the automation can never resolve, so the sweep classifies it as a skipped, non-fatal **unavailable** repository (a `::warning` naming the repository and the remediation) instead of a hard failure. Without this, a handful of un-enrolled repositories keeps the scheduled sweep heartbeat (the org sweep's `0 * * * *` cron) permanently red and masks a genuinely new repository that starts failing. Fail-closed is preserved on both sides: any non-403 scheduler failure still fails the sweep with its per-PR reason, and if more than `ORG_SWEEP_MAX_UNAVAILABLE` (default 5) repositories become unreachable in one pass — a credential-scope regression rather than a few un-enrolled repos — the job fails loudly. Remediation for a listed repository is to install the OpenCode app on it or grant `PR_REVIEW_MERGE_TOKEN` access.
 
 ## Second-reviewer (Noema) posture
 
diff --git a/tests/test_actions_queue_saturation_scheduler_cadence.py b/tests/test_actions_queue_saturation_scheduler_cadence.py
new file mode 100644
index 0000000000..fbf8f45547
--- /dev/null
+++ b/tests/test_actions_queue_saturation_scheduler_cadence.py
@@ -0,0 +1,33 @@
+"""Regression contract for the organization scheduler queue-saturation repair."""
+
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+WORKFLOW = ROOT / ".github" / "workflows" / "pr-review-merge-scheduler.yml"
+
+
+def test_org_queue_sweep_is_hourly_not_quarter_hourly() -> None:
+    """The expensive org sweep must not self-amplify a saturated Actions queue."""
+    workflow = WORKFLOW.read_text(encoding="utf-8")
+    assert '- cron: "0 * * * *"' in workflow or "- cron: '0 * * * *'" in workflow
+    assert '*/15 * * * *' not in workflow
+
+
+def test_org_queue_sweep_wall_clock_fallback_matches_hourly_cadence() -> None:
+    """Fallback rotation and its maintenance comments must match hourly cadence."""
+    workflow = WORKFLOW.read_text(encoding="utf-8")
+    assert workflow.count("$(date -u +%s) / 3600") == 2
+    assert "$(date -u +%s) / 900" not in workflow
+    assert "900s window" not in workflow
+    assert "900s)" not in workflow
+    assert "pending */15 sweep" not in workflow
+
+
+def test_repository_scheduler_keeps_event_driven_wakes() -> None:
+    """Capacity repair must preserve event-driven admission rather than polling only."""
+    workflow = WORKFLOW.read_text(encoding="utf-8")
+    assert "pull_request_target:" in workflow
+    assert "pull_request_review:" in workflow
+    assert "workflow_run:" in workflow
+    assert "repository_dispatch:" in workflow
diff --git a/tests/test_required_workflow_queue_contract.py b/tests/test_required_workflow_queue_contract.py
index 9823c417c1..e18e021fae 100644
--- a/tests/test_required_workflow_queue_contract.py
+++ b/tests/test_required_workflow_queue_contract.py
@@ -967,7 +967,7 @@ def test_org_queue_sweep_covers_target_repositories_on_a_heartbeat() -> None:
     cron, use a cross-repository mutation credential (never the repository
     github.token silently), skip the central repository itself, and fail with a
     visible reason when it cannot mutate sibling repositories. The sweep runs
-    every 15 minutes so an approval that lands after a PR's last event is
+    hourly so an approval that lands after a PR's last event is
     auto-updated/merged promptly instead of idling indefinitely. Its cron has a
     distinct concurrency key from the separate 30-minute scan, and the job has
     enough runtime headroom to finish a complete organization walk.
@@ -975,9 +975,9 @@ def test_org_queue_sweep_covers_target_repositories_on_a_heartbeat() -> None:
     workflow = workflow_text("pr-review-merge-scheduler.yml")
 
     assert "org-queue-sweep:" in workflow
-    assert '- cron: "*/15 * * * *"' in workflow
+    assert '- cron: "0 * * * *"' in workflow
     assert "github.repository == 'ContextualWisdomLab/.github'" in workflow
-    assert "github.event.schedule == '*/15 * * * *'" in workflow
+    assert "github.event.schedule == '0 * * * *'" in workflow
     assert "github.event.client_payload.org_sweep == true" in workflow
     assert (
         "github.event_name == 'schedule' && format('schedule-{0}', "
@@ -994,7 +994,7 @@ def test_org_queue_sweep_covers_target_repositories_on_a_heartbeat() -> None:
     ):
         assert f"{setting}: ${{{{ github.event_name == 'schedule' ||" in workflow
     # The single-repository scan must not double-run on the sweep cron.
-    assert "github.event.schedule != '*/15 * * * *'" in workflow
+    assert "github.event.schedule != '0 * * * *'" in workflow
     assert "github.event.client_payload.org_sweep != true" in workflow
     # The sweep must never silently no-op with the repository-scoped token.
     assert (
@@ -1494,7 +1494,7 @@ def test_org_queue_sweep_treats_inaccessible_repositories_as_non_fatal() -> None
     automation can never resolve, so those repositories are reported as skipped,
     non-fatal "unavailable" repositories rather than hard failures — otherwise a
     handful of un-enrolled repositories keeps the scheduled sweep (the
-    ``*/15 * * * *`` cron) permanently red and masks a genuinely new repository
+    ``0 * * * *`` cron) permanently red and masks a genuinely new repository
     that starts failing.
 
     The sweep stays fail-closed two ways: any non-403 scheduler failure still