Hi, On Wed, Jan 29, 2014 at 5:05 AM, Michael Dürig <[email protected]> wrote: > As mentioned on the issue already I would have waited with that until we > have a at least a better understanding of the limits of the unbounded > approach. In my tests I never saw a significant number of queued events. > Even when sending out millions of events on a heavily loaded system under > memory pressure I never saw more that 1500 queued events. > > While I think this approach is way cool, I also think it unnecessarily > complicates ongoing debugging effort of observation issues we face with > client applications.
I don't see the relevant code as unnecessarily complicated as it's nicely encapsulated in EventGenerator and only kicks in when dealing with huge change-sets. Anyway, I see your point about the benefit being fairly limited, so I wouldn't mind if we dropped or disabled this feature for now. BR, Jukka Zitting
