You could generate tests that give you 100% line coverage but the real test is if the thousands of subscribing players have the same experience which doesn't have the same agent friendly testing feedback loop

What if we replaced the players with AI, also? That would make testing far easier.