This saves:
 - one malloc() per packet per client } due to
 - one memcpy() per packet per client } block_Duplicate()
 - one mwait() per packet per client, and
 - one thread per client per track

So it should really help scale RTSP broadcast up.
However, in its current state, it removes the pts synchronization,
which might not be such a great idea. I will re-add it later.