There is no need to preallocate request buffers for these threads - the event
loops that require them can allocate them when they discover and register the
pre-registered threads. This also means that event loops do not need to
register request buffer factories.