Crowds don't arrive gradually.

When a train berths, a platform's worth of people reach the gateline at once. Sizing that gateline on average throughput guarantees a queue at exactly the moment the queue matters, and the recovery time between services is what decides whether the crowd clears or compounds. Throughput here is a burst figure, not a daily one.

The fare boundary has to hold under that pressure without becoming a safety problem. Gates that separate paid from unpaid area need to release on an alarm or an evacuation signal, immediately and without a credential — the revenue line always loses to the egress route, and the hardware has to be specified so that isn't a decision anyone has to make on the night.

Accessibility is not an edge case in transport. Luggage, prams, bikes, wheelchairs and mobility aids are ordinary traffic, so a wide accessible aisle is part of the standard gateline rather than a token lane at one end.

Finally, these sites are exposed and mostly unattended. Salt air, wind-driven rain, night-time vandalism and continuous operation are the actual duty conditions, and equipment specified for a lobby will not survive them.