I guess XaAES would need its own malloc-lib (one pool for each client) to avoid this "invalid MAGIC"-crap (and maybe improve performance). But that's a long way to go, and before that there are surly some other things that can be improved.
I'm afraid I don't understand the internals of XaAES very well. But IIRC Henk used his malloc-lib which is capable of exactly this when he worked on XaAES. Has this been removed?
Jo Even