> There’s no intrinsic reason it has to be this way other than they built it to be as easy to join as possible.
You are hypothecating a LAMP-style stack where they just need to add a flag to their database table and update all their SQL to ignore data from flagged accounts.
I'm hypothecating architectures where it is substantially more difficult than that, such as heavily cached, event-driven microservices with geographic distribution, data locality, and which attempt to comply with various data retention and privacy regulations.
