Replication, in theory

Experiments are seldom replicated by different research teams, says Philip Ball. Why is this and does it really matter?