I'm no expert, but I can imagine that for sites like Google, Twitter, etc, having a blocklist adds a little complexity, but at their scale, a lot of CPU cycles in overhead. I'm pretty simple in my thought processes though, I'm just thinking that they'd need a unique query for every user, assuming a plain database (works as a mental model). In practice I guess there's one dataset which gets filtered out on its way through the network of servers to the user?