In order to have a reasonable apples-to-apples comparison between the legacy media::AudioOutputController and the new audio::OutputController (for the Audio Service), some existing issues in which "reasons" the stream creation counts and creation times are being recorded were revealed but purposely ignored. This was necessary to resolve  bug 896431  for launch.

For example, OnDeviceChange() will record the change time in the "Media.AudioOutputController.DeviceChangeTime" UMA, but by calling Create(), a UMA there "Media.AudioOutputController.ChangeTime" will be double-tracking the creation time. They really should be separate.

Also, in order to resolve  bug 896431 , a "reason" switch was added for filtering when UMA counts were incremented. However, it seems that this is the wrong filter (but is probably not causing a biased sampling). What really matters is tracking how long it takes to create normal audio streams, regardless of whether that was happening because this was the initial stream creation or because we were toggling muting off.

In all, the counts and timing histograms should be examined to ensure we are not double-tracking anything, and that we are not filtering out things that matter, that we *are* filtering out things that don't matter (e.g., creating fake streams), etc.   ...and this should result in a cleaner code structure as well. :)