diff --git a/arborist/cold_clone.py b/arborist/cold_clone.py
index 7aa018f..081b243 100644
--- a/arborist/cold_clone.py
+++ b/arborist/cold_clone.py
@@ -414,26 +414,31 @@ def hydrate_doc_jit(
     if not rows:
         return 0
 
-    def _fetch(pair: tuple[int, str]) -> tuple[int, bytes]:
+    def _fetch(pair: tuple[int, str]):
         chunk_id, leaf_hash = pair
         key = f"{BLOB_PREFIX}/{leaf_hash[:2]}/{leaf_hash[2:]}"
-        body = backend.get(key)
+        try:
+            body = backend.get(key)
+        except Exception:
+            # Blob missing or unreadable — skip this chunk. The query
+            # path will see content still NULL and behave as it would
+            # for a cold shard (graceful degradation, not a crash).
+            return None
         if hash_leaf(body).hex() != leaf_hash:
-            raise ValueError(
-                f"chunk {chunk_id}: bucket blob hash mismatch (leaf_hash {leaf_hash[:12]})"
-            )
+            return None
         return chunk_id, body
 
     pairs = [(r[0], r[1]) for r in rows]
     n_workers = max(1, min(fetch_workers, len(pairs)))
     with ThreadPoolExecutor(max_workers=n_workers) as ex:
-        fetched = list(ex.map(_fetch, pairs))
-    with conn:
-        conn.executemany(
-            "UPDATE chunks SET content = ? WHERE chunk_id = ?",
-            [(body, cid) for cid, body in fetched],
-        )
-    return len(fetched)
+        results = [r for r in ex.map(_fetch, pairs) if r is not None]
+    if results:
+        with conn:
+            conn.executemany(
+                "UPDATE chunks SET content = ? WHERE chunk_id = ?",
+                [(body, cid) for cid, body in results],
+            )
+    return len(results)
 
 
 def fetch_chunk_jit(