diff --git a/base_tier_validation/__manifest__.py b/base_tier_validation/__manifest__.py
index e826d37b..763d20f6 100644
--- a/base_tier_validation/__manifest__.py
+++ b/base_tier_validation/__manifest__.py
@@ -3,7 +3,7 @@
 {
     "name": "Base Tier Validation",
     "summary": "Implement a validation process based on tiers.",
-    "version": "19.0.1.0.3",
+    "version": "19.0.1.0.4",
     "development_status": "Mature",
     "maintainers": ["LoisRForgeFlow"],
     "category": "Tools",
diff --git a/base_tier_validation/models/tier_review.py b/base_tier_validation/models/tier_review.py
index bed4caaf..00b6c455 100644
--- a/base_tier_validation/models/tier_review.py
+++ b/base_tier_validation/models/tier_review.py
@@ -106,11 +106,28 @@ def _update_review_status(self):
         reviews = self.filtered(lambda rev: rev.status in ["waiting", "pending"])
         if not reviews:
             return
-        next_seq = min(reviews.mapped("sequence"))
+        # Lowest still-open sequence *per record*, computed over the whole
+        # record's reviews -- NOT just this recordset. Computing it over
+        # ``self`` promotes a later tier prematurely whenever the method runs on
+        # a subset that excludes the lower-sequence predecessors, e.g.
+        # ``review_user_count`` calling ``user.review_ids._update_review_status()``
+        # for a second-tier reviewer (their own review is then the only -- and
+        # thus "minimum" -- sequence in the set, so it wrongly goes ``pending``).
+        min_seq_by_record = {}
+        for model, res_id in {(rev.model, rev.res_id) for rev in reviews}:
+            open_reviews = (
+                self.env[model]
+                .browse(res_id)
+                .review_ids.filtered(lambda r: r.status in ("waiting", "pending"))
+            )
+            min_seq_by_record[(model, res_id)] = min(open_reviews.mapped("sequence"))
         for record in reviews:
             if record.status != "waiting":
                 continue
-            if record.approve_sequence and record.sequence != next_seq:
+            if (
+                record.approve_sequence
+                and record.sequence != min_seq_by_record[(record.model, record.res_id)]
+            ):
                 continue
             record.status = "pending"
             if record.definition_id.notify_on_pending:
diff --git a/base_tier_validation/tests/test_tier_validation.py b/base_tier_validation/tests/test_tier_validation.py
index 3f2ab80c..3414b21b 100644
--- a/base_tier_validation/tests/test_tier_validation.py
+++ b/base_tier_validation/tests/test_tier_validation.py
@@ -689,6 +689,67 @@ def test_19a_notify_on_pending_sequence_negative(self):
             "Second-tier reviewer must be notified once promoted to pending.",
         )
 
+    def test_19c_no_premature_promotion_on_subset(self):
+        """Regression: ``_update_review_status`` run on a *subset* of a record's
+        reviews -- as ``review_user_count`` does per user, via
+        ``user.review_ids._update_review_status()`` -- must gate the sequence
+        over the *whole* record, not the subset. Otherwise a later-tier review
+        is promoted to ``pending`` before its predecessor is approved, which
+        both hides it from the reviewer's systray (``can_review`` is False while
+        the lower tier is still pending) and fires a ``notify_on_pending``
+        message authored by -- and so never delivered to -- that reviewer.
+        """
+        TierDefinition = self.env["tier.definition"]
+        test_record = self.test_model.create({"test_field": 2.5})
+        TierDefinition.create(
+            {
+                "model_id": self.tester_model.id,
+                "review_type": "individual",
+                "reviewer_id": self.test_user_1.id,
+                "definition_domain": "[('test_field', '=', 2.5)]",
+                "approve_sequence": True,
+                "notify_on_pending": True,
+                "sequence": 20,
+                "name": "First in sequence -- user 1",
+            }
+        )
+        TierDefinition.create(
+            {
+                "model_id": self.tester_model.id,
+                "review_type": "individual",
+                "reviewer_id": self.test_user_2.id,
+                "definition_domain": "[('test_field', '=', 2.5)]",
+                "approve_sequence": True,
+                "notify_on_pending": True,
+                "sequence": 10,
+                "name": "Second in sequence -- user 2",
+            }
+        )
+        reviews = test_record.request_validation()
+        review_second = reviews.filtered(lambda r: self.test_user_2 in r.reviewer_ids)
+        self.assertEqual(review_second.status, "waiting")
+        messages_before = test_record.message_ids
+
+        # Simulate the second-tier reviewer opening their systray: the promotion
+        # runs on *only* their own review -- the subset that excludes the
+        # first-tier predecessor.
+        review_second._update_review_status()
+        review_second.invalidate_recordset()
+
+        self.assertEqual(
+            review_second.status,
+            "waiting",
+            "A later-tier review must stay 'waiting' until its predecessor is "
+            "approved, even when _update_review_status runs on a subset that "
+            "excludes the predecessor.",
+        )
+        self.assertEqual(
+            test_record.message_ids,
+            messages_before,
+            "No notify_on_pending message must be posted from promoting a later "
+            "tier out of order.",
+        )
+
     def test_19b_notify_review_available_no_op_when_no_users(self):
         """``_notify_review_available`` must short-circuit (no follower
         added, no chatter message posted) when none of the passed reviews