The user probably has code objects (stored proc/pkgs) as well owns tables
which are being replicated elsewhere? The lib cache pins can be explained by
the need to lock/pin affected objects in the shared pool for invalidation.
These objects are code owned by the user or are referring the objects own
I'm doing some departed user cleanup and the 'drop user cascade' hangs
for 2 -3 minutes before completeing. I ran a 10046 trace and it does a lot
of waiting for 'library cache pin' - 100 times for 308 centiseconds each
time. Dropped two different users, 100 library cache pin waits per user.