diff --git a/rq/contrib/legacy.py b/rq/contrib/legacy.py
new file mode 100644
index 0000000..2a20b9f
--- /dev/null
+++ b/rq/contrib/legacy.py
@@ -0,0 +1,24 @@
+import logging
+from rq import get_current_connection
+from rq import Worker
+
+
+logger = logging.getLogger(__name__)
+
+
+def cleanup_ghosts():
+    """
+    RQ versions < 0.3.6 suffered from a race condition where workers, when
+    abruptly terminated, did not have a chance to clean up their worker
+    registration, leading to reports of ghosted workers in `rqinfo`.  Since
+    0.3.6, new worker registrations automatically expire, and the worker will
+    make sure to refresh the registrations as long as it's alive.
+
+    This function will clean up any of such legacy ghosted workers.
+    """
+    conn = get_current_connection()
+    for worker in Worker.all():
+        if conn._ttl(worker.key) == -1:
+            ttl = worker.default_worker_ttl
+            conn.expire(worker.key, ttl)
+            logger.info('Marked ghosted worker {0} to expire in {1} seconds.'.format(worker.name, ttl))
diff --git a/rq/scripts/rqworker.py b/rq/scripts/rqworker.py
index 28f3921..61ecbb0 100755
--- a/rq/scripts/rqworker.py
+++ b/rq/scripts/rqworker.py
@@ -7,6 +7,7 @@ import logging.config
 from rq import Queue, Worker
 from rq.logutils import setup_loghandlers
 from redis.exceptions import ConnectionError
+from rq.contrib.legacy import cleanup_ghosts
 from rq.scripts import add_standard_arguments
 from rq.scripts import setup_redis
 from rq.scripts import read_config_file
@@ -48,6 +49,8 @@ def main():
     setup_loghandlers(args.verbose)
     setup_redis(args)
 
+    cleanup_ghosts()
+
     try:
         queues = map(Queue, args.queues)
         w = Worker(queues, name=args.name)