It's mainly to provide discrimination between protective devices, partly to do with cost and very slightly to do with safety.
Most circuits above 32A are used for distribution to final circuits. As such it is preferable for a fault occuring on a final circuit, not to cause the protective device for the distribution circuit to operate.
Some circuits above 32A are final circuits, but as they are generally used to supply fixed equipment, the likelyhood of damage occurring to flexible leads is slight, and the cost of providing a 0.4s disconnection time prohibitive. Also many of these circuits are required to be additionally protected by a 30mA RCD, so the disconnection time becomes rather moot.
It used to be in the 16th, that the 0.4s disconnection time only applied to circuits supplying socket-outlets or other circuits supplying hand held or portable equipment.
Obviously a 0.4s disconnection time is safer than a 5s disconnection time. Lighting circuits were allowed a 5s disconnection time.
The 0.4s disconnection time was extended to cover lighting circuits in the 17th, to provide greater safety, probably because of the increase in the use of metal light switches, and the increased use of socket-outlets for connecting standard lamps and table lamps etc.