I am consider using fly.io postgres for a highly replicated KV store that will push down changes to a high number of regions. Sort of like Cloudflare KV but pushing updates down to the cached regions rather than waiting for the first request and getting at least one very slow response.
Are there any limits in terms of number of replicas that I should be considering for this? For example if I made a primary instance with 8vCPU and then had a 4vCPU replica in EVERY available region, what problems start to arise?
This is like 99.999999% read operations so the optimization is for lowering read latency around the world, without waiting for a request to cache a value locally.