JFR: onFlush invoked out of order with a sorted event stream

The purpose of the onFlush handler is to allow clients to redraw a graph in a GUI or push event data over the network when a segment of the file is complete. When a stream is sorted and the parser is behind, which always happens when opening a stream against a file, the events are aggregated over several segments and sent out as a batch.

The flush handlers are on the other hand invoked when they are discovered on disk. This can results in hundreds of calls to onFlush when the recording starts and before even one single event is sent. This also mean the onFlush handler is not called when a file ends.

Impact: High, potential loss of data since client will not get onFlush call at the end.
Likelihood: Low, only happens when parser is behind, or when streaming against a file
Workaround: High, no known workaround

HLH = > P2