As for audio data stored in banks, it is copied in a separate pool for each LoadBank() calls, or in a user-defined pool, giving you great control on memory management of audio data.
The memory allocator used by the Wwise sound engine is extremely efficient in regard to fragmentation. Normally, memory usage should not build up significantly due to fragmentation. However, the Lower Engine Default pool may suffer from such fragmentation when it doesn't have enough headroom above its typical usage: some buffers in this pool may be very big, like reverbs with large decay times. Obviously, a big delay line needs to be allocated in one contiguous chunk.Please read http://kb.gowwise.com/questions/152/Determining+an+Appropriate+Size+for+Your+Memory+Pools for tips on determining appropriate sizes for the memory pools.
