Well, aggr requires AUTH, right, so you should only be serving them up to the original pubkey. But that means they author can't use aggr to see the rest of the "-" thread, so it's not really useful.
Discussion
Will fix eventually. Currently busy doing distributed systems
Yeah, no rush. It's public data.
Working on cursor queries currently. Instead of having to do created_at based pagination which can be problematic, you just pass the cursor in and resume where it left off.
Works across connections too.
May implement reverse queries as well as it is minimal effort