One could create multiple apps (each get their own IP), and plonk the IPs in any number of DNS A
/AAAA
/HTTPS
/SVCB
records as you see fit (1).
One could run one server per region, assign them regional IPs, and map those to DNS records (2).
One can run app over multiple Machines (which, for now, each get a fixed public IPv6 address) (3).
Anchor scaling also helps keep public IPv6 stable for regular Fly app instances, but I see references to it have been removed from docs. Anyways, ref: (4).
One could route traffic to a set of front-end app instances that then forward it to appropriate back-end app instance: