diff -u solr/core/src/java/org/apache/solr/search/SolrIndexSearcher.java solr/core/src/java/org/apache/solr/search/SolrIndexSearcher.java
--- solr/core/src/java/org/apache/solr/search/SolrIndexSearcher.java	(working copy)
+++ solr/core/src/java/org/apache/solr/search/SolrIndexSearcher.java	(working copy)
@@ -178,8 +178,6 @@
 
     fieldNames = r.getFieldNames(IndexReader.FieldOption.ALL);
 
-    // TODO: don't do this until its safe
-    setFilterRandomAccessThreshold(0);
     // do this at the end since an exception in the constructor means we won't close    
     numOpens.incrementAndGet();
   }
diff -u lucene/src/java/org/apache/lucene/search/FilteredQuery.java lucene/src/java/org/apache/lucene/search/FilteredQuery.java
--- lucene/src/java/org/apache/lucene/search/FilteredQuery.java	(working copy)
+++ lucene/src/java/org/apache/lucene/search/FilteredQuery.java	(working copy)
@@ -116,6 +116,10 @@
         }
 
         final int firstFilterDoc = filterIter.nextDoc();
+        if (firstFilterDoc == DocIdSetIterator.NO_MORE_DOCS) {
+          return null;
+        }
+        
         final Bits filterAcceptDocs = filterDocIdSet.bits();
         final boolean useRandomAccess = (filterAcceptDocs != null && firstFilterDoc < searcher.getFilterRandomAccessThreshold());
 
diff -u lucene/src/java/org/apache/lucene/search/DocIdSet.java lucene/src/java/org/apache/lucene/search/DocIdSet.java
--- lucene/src/java/org/apache/lucene/search/DocIdSet.java	(working copy)
+++ lucene/src/java/org/apache/lucene/search/DocIdSet.java	(working copy)
@@ -61,9 +61,19 @@
    * are no docs that match. */
   public abstract DocIdSetIterator iterator() throws IOException;
 
-  /** Optionally provides a {@link Bits} interface for random access.
-   * Returns {@code null}, if this DocIdSet does not support random access.
-   * The default implementation does not provide random access */
+  /** Optionally provides a {@link Bits} interface for random access
+   * to matching documents.
+   * @return {@code null}, if this {@code DocIdSet} does not support random access.
+   * In contrast to {@link #iterator()}, a return value of {@code null}
+   * <b>does not</b> imply that no documents match the filter!
+   * The default implementation does not provide random access, so you
+   * only need to implement this method if your DocIdSet can
+   * guarantee random access to every docid in O(1) time without
+   * external disk access (as {@link Bits} interface cannot throw
+   * {@link IOException}). This is generally true for bit sets
+   * like {@link org.apache.lucene.util.FixedBitSet}, which return
+   * itsself if they are used as {@code DocIdSet}.
+   */
   public Bits bits() throws IOException {
     return null;
   }