There is an existing Shopify connector, so I'd recommend using separate source and connection for every store. You can use Airbyte terraform provider to configure all connections. terraform has for_each loop, so you iterate over set of credentials to configure your sources & connections. With Kubernetes deployment and autoscaling you can have many connections running in parallel.
Putting so many sources into one connection feels like a huge bottleneck.
If you write your own connector, are you going to run 500 threads or processes in single Docker container? Probably not, so data synchronization will be running only for smaller part of those stores.
What will happen in case of any network issues with one of those stores? Or corrupted state?
How long will it take to synchronize the data from all 500 stores?