This paper discusses the role and implementation of asynchronous event handlers in the real-time specification for Java (RTSJ). For non-blocking handlers, an implementation model whereby all heap-using handlers are serviced by a single thread and all no-heap handlers are serviced by another server thread is proposed and schedulability analysis is derived. The model is shown to have bounded priority inversion. General multiple-server models are needed for nonblocking handlers but the support in the RTSJ is criticised as lacking in configurability. A solution is proposed which allows the number of servers to be specified, and the allocation of handlers to servers to be controlled.