Split off from {T182314}:
When analyzing the data from the [[https://meta.wikimedia.org/wiki/Schema:Popups | Popups schema]] about the duration of user interactions with the page previews feature (from the begin of a mouse hover over a link to the user action occurring), we encountered strange patterns such as this:
{F13248348,width=100%}
(from T182314#3954765, see there for data source)
See the discussion at T182314 for what we found out about this so far, e.g. the detailed explanations about timers by @phuedx in T182314#3956099 and subsequent observations by @Jhernandez, @Gilles and others. While in that investigation we already figured out a lot of things about how the underlying timers work and reached a level of confidence that these anomalies should not materially affect the main takeaways for that A/B test, we have not yet found a satisfying explanation for them.
Since this kind of thing is likely to occur in similar web instrumentations in the future, there has been a sense that it warrants further investigation. In particular we would want to rule out the possibility that it is a "data smell" indicative of deeper problems with this method of measuring interaction times, rather than just a limitation of its accuracy.