diff --git a/tests/test_sync_error_classification.py b/tests/test_sync_error_classification.py
index 391a8f0..dc4bff3 100644
--- a/tests/test_sync_error_classification.py
+++ b/tests/test_sync_error_classification.py
@@ -43,9 +43,42 @@ class SyncErrorClassificationTest(unittest.TestCase):
 
         self.assertEqual(
             [(value["generation_id"], value["result"]) for value in acknowledgements],
-            [("gen_consumed", "consumed"), ("gen_rejected", "rejected")],
+            [("gen_rejected", "rejected"), ("gen_consumed", "consumed")],
         )
 
+    def test_recovery_envelope_prioritizes_recent_terminal_evidence(self):
+        with sqlite3.connect(":memory:") as database:
+            database.execute(
+                "CREATE TABLE sync_consumed_generations("
+                "generation_id TEXT,producer_peer_id TEXT,consumer_peer_id TEXT,"
+                "consumed_at TEXT,result TEXT,ack_json TEXT)"
+            )
+            for index in range(40):
+                generation = f"gen_{index:02d}"
+                database.execute(
+                    "INSERT INTO sync_consumed_generations VALUES(?,?,?,?,?,?)",
+                    (
+                        generation,
+                        "android",
+                        "desktop",
+                        f"2026-09-18T12:{index:02d}:00Z",
+                        "consumed",
+                        json.dumps({
+                            "generation_id": generation,
+                            "result": "consumed",
+                            "diagnostic": "",
+                        }),
+                    ),
+                )
+
+            acknowledgements = sync_peer_worker.recovery_acknowledgements(
+                database, "android", "desktop"
+            )
+
+        self.assertEqual(32, len(acknowledgements))
+        self.assertEqual("gen_39", acknowledgements[0]["generation_id"])
+        self.assertEqual("gen_08", acknowledgements[-1]["generation_id"])
+
     def test_mtp_refreshes_a_retained_peer_advertisement_before_validation(self):
         expected_peer = "peer_00000000-0000-4000-8000-000000000001"
         with tempfile.TemporaryDirectory() as directory:
diff --git a/tools/sync_peer_worker.py b/tools/sync_peer_worker.py
index c1f3dc5..e041caf 100644
--- a/tools/sync_peer_worker.py
+++ b/tools/sync_peer_worker.py
@@ -72,10 +72,14 @@ def recovery_acknowledgements(db, producer_peer_id: str, consumer_peer_id: str)
     rows = db.execute(
         "SELECT ack_json FROM sync_consumed_generations WHERE producer_peer_id=? "
         "AND consumer_peer_id=? AND result IN('consumed','rejected') "
-        "ORDER BY consumed_at,generation_id LIMIT 32",
+        "ORDER BY consumed_at DESC,generation_id DESC LIMIT 32",
         (producer_peer_id, consumer_peer_id),
     ).fetchall()
     acknowledgements = [json.loads(row[0]) for row in rows]
+    # WHY: the bounded envelope must prioritize the newest interrupted
+    # conversations. Older acknowledged generations are normally already
+    # archived; selecting them first can starve the exact ACKs needed to free
+    # active producer capacity.
     # ACKs created before rejection diagnostics were normalized may contain a
     # solidus that Android's platform JSON canonicalizer hashes differently.
     # They remain durable evidence but cannot close a peer row, so do not let