Strange replication behaviours

[Jochen Kempf]

Please see http://gist.github.com/380150
(s_p_a_m check doesn't allow me to send directly to the list)