It is desirable keep have high mutator utilization (where utility is the fraction of time that the mutator gets to itself in some fixed window of time). We don't know right now what the utilization is; our instrumentation reports pause data, but we don't know much about how those pauses are clustered, so one cannot conclude from a low maximum pause time that the minimum mutator utllization (MMU) is high. Measuring utilization inherently requires history during a run. Simplest thing in short term would be to log a timestamp and tag to an internal buffer at each phase shift and then dump it all out at the end of the run. (If we can layer this atop the Sampler, all the better.) Then post-process the result with an external tool to extract utilization information. Longer term we may need to do something smarter. If it comes to that, we'll log a separate bug/task entry for it.
More background and general overview on Zerowing at: [FlashPlayer:MMgc Metrics - Goals and Status] https://zerowing.corp.adobe.com/x/TAX4H
(very sad that i never got to even prototype this.)