It would be good to look at how many of these numbers are negative.
That seems right. I think there's a real problem here, since physical clocks do just fail, and sometimes NTP slew has to happen. LoopingCall.withCount has a clear opportunity to notice that time has gone backwards, and can just skip that call. This is assuming that time has gone backwards by a tiny amount for a brief period of time, though, and does not address the more complex issue of "the clock was set some significant distance into the past". LoopingCall still isn't going to be able to deal with large amounts of clock drift, we'd need a monotonic version for that :).I must add that the anomaly doesn't appear right at the start, but after tens of seconds, mostly when the system is under medium/heavy load.Even seriously broken clocks don't go backwards a lot. ;-)
Thanks for reporting! Can you open a ticket?