Hi,
The difficulty is that the deployment load balancer will stop you being able to communicate directly with each gateway to configure it. You could get around this by opening a shell in each gateway pod and configuring the api and policy definitions via the localhost interface then calling the reload API.
The other option would be to have multiple namespaces with individual gateways in each. That would allow you to use the operator in each namespace to configure each gateway, and then have another higher level load balancer to spread the traffic between gateways.
But don’t underestimate the amount of traffic that a single gateway can handle. It’s CPU bound so if provided with enough CPU an single gateway can handle a lot of traffic. Please see Planning for Production
Cheers,
Pete