I would try turning on analysis again with our next release, and reporting back. We’ve implemented some much more aggressive buffering to address some scattered reports of dropouts with certain configurations. Would be interesting to know if it helps.