diff --git a/worker/consumer/consumer.py b/worker/consumer/consumer.py
index 0aecbb9..a23eb30 100644
--- a/worker/consumer/consumer.py
+++ b/worker/consumer/consumer.py
@@ -27,6 +27,7 @@
 
 from __future__ import annotations
 
+import time
 from typing import Dict
 
 import redis
@@ -84,6 +85,13 @@ def __init__(
         self.storage = storage
         self.cfg = cfg
 
+        # Periodic recovery state. _last_recovery = 0 makes recovery run on the
+        # first consume() so leftovers from a prior crash are swept at startup.
+        # The interval matches the 2-minute staleness threshold in the recovery
+        # query. See DEV-35.
+        self._last_recovery = 0.0
+        self._recovery_interval = 120.0
+
         # Ensure the consumer group exists. If it already exists Redis raises an
         # error; ignore that specific error.
         try:
@@ -121,6 +129,11 @@ def consume(self, consumer_name: str) -> bool:
             True if a message was consumed (even if processing failed), False if
             no messages were available.
         """
+        # Recover stuck jobs on a fixed cadence, independent of load. Doing this
+        # only on the idle path meant recovery never ran under sustained load —
+        # exactly when crashed-mid-job rows are most likely. See DEV-35.
+        self._maybe_recover()
+
         # Read one message for this consumer (blocking short period)
         resp = self.redis.xreadgroup(
             groupname=self.cfg.consumer_group,
@@ -131,8 +144,6 @@ def consume(self, consumer_name: str) -> bool:
         )
 
         if not resp:
-            # No messages available; attempt recovery of stuck jobs and return.
-            self._recover_stuck_pending()
             return False
 
         # Response format: [(stream_name, [(msg_id, {field: value}), ...])]
@@ -314,6 +325,17 @@ def _handle_asset_message(self, asset_id: str, msg_id: str) -> None:
         # Delegate to _handle_job using the job id we now have.
         self._handle_job(job_id, msg_id)
 
+    def _maybe_recover(self) -> None:
+        """Run stuck-job recovery if the recovery interval has elapsed.
+
+        Time-gated so recovery fires on a fixed cadence regardless of whether
+        the consumer is busy or idle.
+        """
+        now = time.time()
+        if now - self._last_recovery >= self._recovery_interval:
+            self._last_recovery = now
+            self._recover_stuck_pending()
+
     def _recover_stuck_pending(self) -> None:
         """Requeue stale pending/in_progress jobs back onto the stream.
 
diff --git a/worker/tests/test_consumer_recovery.py b/worker/tests/test_consumer_recovery.py
new file mode 100644
index 0000000..d76cb40
--- /dev/null
+++ b/worker/tests/test_consumer_recovery.py
@@ -0,0 +1,56 @@
+import time
+import unittest
+from unittest.mock import MagicMock, patch
+
+from worker.consumer.consumer import Consumer
+
+
+def _make_consumer():
+    """Build a Consumer with redis/pg mocked out (no real connections)."""
+    cfg = MagicMock()
+    cfg.stream_name = "media:jobs"
+    cfg.consumer_group = "media-workers"
+    with patch("worker.consumer.consumer.redis.Redis.from_url") as from_url:
+        client = MagicMock()
+        from_url.return_value = client
+        consumer = Consumer(
+            pg_pool=MagicMock(), redis_url="redis://x", storage=MagicMock(), cfg=cfg
+        )
+    return consumer, client
+
+
+class TestPeriodicRecovery(unittest.TestCase):
+    """DEV-35: recovery must fire on a cadence even under continuous load."""
+
+    def test_recovery_fires_under_load_when_interval_elapsed(self):
+        consumer, client = _make_consumer()
+        # A message is available (the loaded / busy path).
+        client.xreadgroup.return_value = [("media:jobs", [("1-0", {"job_id": "42"})])]
+
+        consumer._recover_stuck_pending = MagicMock()
+        consumer._handle_job = MagicMock()
+        consumer._recovery_interval = 0.0  # gate always open
+        consumer._last_recovery = 0.0
+
+        result = consumer.consume("worker-1")
+
+        self.assertTrue(result)  # work was performed
+        consumer._handle_job.assert_called_once_with("42", "1-0")
+        consumer._recover_stuck_pending.assert_called_once()  # fired despite load
+
+    def test_recovery_does_not_fire_before_interval_elapses(self):
+        consumer, client = _make_consumer()
+        client.xreadgroup.return_value = [("media:jobs", [("1-0", {"job_id": "42"})])]
+
+        consumer._recover_stuck_pending = MagicMock()
+        consumer._handle_job = MagicMock()
+        consumer._recovery_interval = 9999.0
+        consumer._last_recovery = time.time()  # just recovered
+
+        consumer.consume("worker-1")
+
+        consumer._recover_stuck_pending.assert_not_called()  # gate holds
+
+
+if __name__ == "__main__":
+    unittest.main()