To follow up on that one,

In my experiments, the 'nightmare' case (one physical page benig mapped
at least twice within the same process) never happened. Could you
comment on when this does actually happen in practice?

	Neil