Nexon's decline in traffic

Additional frame data has zero overhead in the files. Actually, having the frame data stored locally in each img file would be faster than using a link because there's no redirection involved.

I don't entirely agree with you on that one, but it largely depends on implementation.

A pointer is a pointer regardless, whether you're pointing from vnum->img[0] or vnum->link->img[0] the only real point where one would be more cumbersome than the other is in establishing the link pointer itself, which would normally be a one time load during client initialization.

I do agree having redundant images wouldn't matter much because they're nested, and therefore wouldn't impede traversal of anything but themselves, and the actual space they occupy in the files. They don't necessarily have to occupy any additional memory because a clever programmer could easily store images with identical checksums in a hash table and share references if it were truly a concern.

An event handler with lazy loading would be enough to solve that issue. If no events have ever been reported, then that place should never be polled. Only load the resource if an event is reported from that resource.

You could also do this by traversing a linked list of active areas only and moving the areas to/from the active pool based on inactivity. A lot of possible solutions but the only one that matters is whatever they actually did