Python 2.7 tests were too slow.

For some reason (I have not investigated) the config loading
takes longer in python 2.7 than in python 3. This caused a
higher probability for the watching tests to fail.

tests/spec_test.py

@@ -869,7 +869,7 @@ def change_config(label):
 
     wait_time = 0.0
 
-    while any(flags.values()) and wait_time <= 3:
+    while any(flags.values()) and wait_time <= 4:
         time.sleep(0.25)
         wait_time += 0.25