diff --git a/setup.py b/setup.py
index 40a5d9d..9ad4da4 100755
--- a/setup.py
+++ b/setup.py
@@ -50,7 +50,7 @@ except:
 
 setuptools.setup(
         name='warcprox',
-        version='2.1b1.dev90',
+        version='2.1b1.dev91',
         description='WARC writing MITM HTTP/S proxy',
         url='https://github.com/internetarchive/warcprox',
         author='Noah Levitt',
diff --git a/tests/test_warcprox.py b/tests/test_warcprox.py
index dd80a86..55e655f 100755
--- a/tests/test_warcprox.py
+++ b/tests/test_warcprox.py
@@ -1429,6 +1429,45 @@ def test_via_response_header(warcprox_, http_daemon, archiving_proxies, playback
                 elif record.rec_type == 'request':
                     assert record.http_headers.get_header('via') == '1.1 warcprox'
 
+def test_summy_merge():
+    d1 = {
+        'a': {
+            'metadata': 'some value',
+            'a1': 5,
+            'a2': 6,
+        },
+        'b': {
+            'b1': 9,
+        }
+    }
+
+    d2 = {
+        'a': {
+            'a1': 7,
+            'a3': 8,
+        },
+        'c': {
+            'c1': 10,
+        }
+    }
+
+    merged = {
+        'a': {
+            'metadata': 'some value',
+            'a1': 12,
+            'a2': 6,
+            'a3': 8,
+        },
+        'b': {
+            'b1': 9,
+        },
+        'c': {
+            'c1': 10,
+        },
+    }
+
+    assert warcprox.stats.summy_merge(d1, d2) == merged
+
 if __name__ == '__main__':
     pytest.main()
 
diff --git a/warcprox/stats.py b/warcprox/stats.py
index 6e0afdd..682d197 100644
--- a/warcprox/stats.py
+++ b/warcprox/stats.py
@@ -32,6 +32,7 @@ import rethinkdb as r
 import datetime
 import urlcanon
 import sqlite3
+import copy
 
 def _empty_bucket(bucket):
     return {
@@ -50,6 +51,28 @@ def _empty_bucket(bucket):
         },
     }
 
+def summy_merge(e, f):
+    if isinstance(e, (int, float)) and isinstance(f, (int, float)):
+        return e + f
+    elif (e is not None and not hasattr(e, 'keys')) or (
+            f is not None and not hasattr(f, 'keys')):
+        return e or f
+    else:
+        result = {}
+        all_keys = set(e.keys()).union(f.keys())
+        for k in all_keys:
+            m = e.get(k)
+            n = f.get(k)
+            if m is None and isinstance(n, (int, float)):
+                m = 0
+            elif n is None and isinstance(m, (int, float)):
+                n = 0
+            else:
+                m = m or {}
+                n = n or {}
+            result[k] = summy_merge(m, n)
+        return result
+
 class StatsDb:
     logger = logging.getLogger("warcprox.stats.StatsDb")
 
@@ -207,45 +230,51 @@ class RethinkStatsDb(StatsDb):
         """Starts batch update repeating timer."""
         self._update_batch() # starts repeating timer
 
-    def _bucket_batch_update_reql(self, bucket):
+    def _bucket_batch_update_reql(self, bucket, batch):
         return self.rr.table(self.table).get(bucket).replace(
             lambda old: r.branch(
-                old.eq(None), self._batch[bucket], old.merge({
+                old.eq(None), batch[bucket], old.merge({
                     "total": {
                         "urls": old["total"]["urls"].add(
-                            self._batch[bucket]["total"]["urls"]),
+                            batch[bucket]["total"]["urls"]),
                         "wire_bytes": old["total"]["wire_bytes"].add(
-                            self._batch[bucket]["total"]["wire_bytes"]),
+                            batch[bucket]["total"]["wire_bytes"]),
                         },
                     "new": {
                         "urls": old["new"]["urls"].add(
-                            self._batch[bucket]["new"]["urls"]),
+                            batch[bucket]["new"]["urls"]),
                         "wire_bytes": old["new"]["wire_bytes"].add(
-                            self._batch[bucket]["new"]["wire_bytes"]),
+                            batch[bucket]["new"]["wire_bytes"]),
                         },
                     "revisit": {
                         "urls": old["revisit"]["urls"].add(
-                            self._batch[bucket]["revisit"]["urls"]),
+                            batch[bucket]["revisit"]["urls"]),
                         "wire_bytes": old["revisit"]["wire_bytes"].add(
-                            self._batch[bucket]["revisit"]["wire_bytes"]),
+                            batch[bucket]["revisit"]["wire_bytes"]),
                         },
                 })))
 
     def _update_batch(self):
+        with self._batch_lock:
+            batch_copy = copy.deepcopy(self._batch)
+            self._batch = {}
         try:
-            with self._batch_lock:
-                if len(self._batch) > 0:
-                    # XXX can all the buckets be done in one query?
-                    for bucket in self._batch:
-                        result = self._bucket_batch_update_reql(bucket).run()
-                        if (not result["inserted"] and not result["replaced"]
-                                or sorted(result.values()) != [0,0,0,0,0,1]):
-                            raise Exception(
-                                    "unexpected result %s updating stats %s" % (
-                                        result, self._batch[bucket]))
-                    self._batch = {}
+            if len(batch_copy) > 0:
+                # XXX can all the buckets be done in one query?
+                for bucket in batch_copy:
+                    result = self._bucket_batch_update_reql(
+                            bucket, batch_copy).run()
+                    if (not result["inserted"] and not result["replaced"]
+                            or sorted(result.values()) != [0,0,0,0,0,1]):
+                        raise Exception(
+                                "unexpected result %s updating stats %s" % (
+                                    result, batch_copy[bucket]))
         except Exception as e:
             self.logger.error("problem updating stats", exc_info=True)
+            # now we need to restore the stats that didn't get saved to the
+            # batch so that they are saved in the next call to _update_batch()
+            with self._batch_lock:
+                self._batch = summy_merge(self._batch, batch_copy)
         finally:
            if not self._stop.is_set():
                self._timer = threading.Timer(2.0, self._update_batch)