maybe it would be a good idea to add a flag for skipping EventHandlers, like Actor.registry.shutdownAll(skipLoggers = false)

I don't think we should add special things for the event handler listeners in the api. They are like any other actor, and application specific listeners can be added in runtime to the EventHandler. There is foreach and filter method in ActorRegistry so it would be easy to do special shutdown if needed.

I think the tricky part of this is how do we know that an actor is an event handler listener?

I have an idea, will try it on 1.2 first.

See last two commits on testkit branch: I introduced a generic shutdownLevel to Actor/ActorRegistry and used that to configure the EventHandler.DefaultListener to be stopped late (user configurable). This should then be applied to all other actors started/used by the toolkit, if the method is accepted.

Updating tickets (#841, #844, #859, #862, #897, #898, #899, #900, #901, #903, #905)

This ticket was closed prematurely (by myself, I admit). The fix was not accepted by Viktor and thus not integrated into the release branch.

I'm not really a huge fan of shutdownAll... should be more like "killAll" so people don't expect any ordering or otherwise, should be a crash-only thing, right?

Updating tickets (#859, #1036, #1037, #1067, #1085, #1105, #1106, #1166)

fixed on logging branch: system services (to which configured loggers belong) are stopped after stopping all normal actors and then re-attaching the primitive stdout-logger so that even the shutdown of loggers may do logging.