Description:
A feature request to have some way to find out where the server is allocating RAM so it's possible to track down why it's using more than expected. Both for general allocations and during query execution. Note also that storage engines are often the big users of RAM so this includes finding some way for them to report their usage.
How to repeat:
find a server using mroe RAM than expected, try to work out where it's allocating the RAM and find out that there's no way to get a total RAM report so you can track it down.
Suggested fix:
Looks like work for the information schema mainly. Query execution/parsing might be suitable for explain.