I was getting intermittent test crashes, I tracked this
to the fact that the JDI VMDeathEvent effectively has a suspend
policy of SUSPEND_NONE, it was originally intended to have
a SUSPEND_ALL policy. Result is there is no way to safely sync
SUSPEND_NONE events before VM termination; That is, there is
no way to be sure that all the events in the event stream are
processed before the connection goes away; This is not a
problem for SUSPEND_ALL events since they sync themselves.

it is also a problem for any client debugger with asynchronous
operation, since there is no way to prevent operations after
VMDeath and thus also no way to prevent operations after
VMDisconnect.