Yes, you need to create a separate database for each LoRa Server instance to separate the data between these network-servers. Please note that these these databases can exist on the same PostgreSQL instance (e.g. loraserver_ns_band_a, loraserver_ns_band_b, …).
But this begs the question that why are we using such an architecture? IMHO this increases the number of managed instances which I was trying to avoid.
Since every region has different settings which can be customized. Some regions cover multiple countries where you want to make per-country modifications to these settings.
It also allows you to run the India instance at a datacenter close to / in India, same for the European instance. That way you’ll be able to keep the latency between the network-servers and the gateways low 