Blocking send functions. If no outbound queue space is available they block until either space becomes available or a GENEROUS time limit is exceeded, in which case it panics diagnostically. This relieves the caller of the burden of supporting an error case which will only be seen in a grossly malfunctioning system and so will be hard to test and difficult to recover from. To avoid deadlock due to the peer being in the same state the inbound queue is drained while waiting for space. The receipt order is always preserved.

The caller intends to block for some out-of-band synchronisation mechanism (eg a semaphore) and so dispatch must be done immediately as part of this sending call if the peer object is running in the same worker thread, so that the recipient can signal the caller if appropriate. Such synchronisation mechanisms are strongly discouraged but exist in some legacy code.