In some dispatching strategies the event service uses one or more queues
serviced by dispatching
threads.  The maximum size of said queues is currently hard-coded.  It should be
selectable with
at least the following options:
- No limit
- Max number of messages
- Max total size (can we measure that?)

once the maximum is reached the queue simply flows controls the suppliers.  In
some applications
we may wish to drop events, for example:
- The oldest events are dropped to make room for the new one.
- or the the events with lowest priority

this last aspect should probably be controlled using some application specific
strategy,
because there may be multiple different criteria to drop the events.