I have two ideas to combat #spam attacks like the one happening now. These are both on the client side.
1) The client could mark posts that mention the user as spam, if the sender is not followed by the user or the sender does not follow the user (this could be a setting). Spam mentions would be collected in their own area separate from normal notifications.
2) The client could group notifications for posts which share most of the same body text, into a single notification.