This is just a thought. Trying to spawn some ideas.

This request is to replicate the Query Cache, one instance for each table. This would allow for a slightly more fine-grained control of the query cache. So, instead of storing the name of the table involved with each query, each table would have a reference to its own instance of the Query Cache.

For queries that involved more than one table, a "general" Query Cache would be employed that acts much like the current implementation does.

An immediate downside I can see is that each time a query comes it, it would have to do a linear search of the packet in all the caches.

Because each table has its own cache, any cache could be disabled. This would have the effect of ignoring (not caching) queries for a given a table.