We are working on some design updates and I’ll see if we can integrate some of these ideas.
I’m concerned with this behaviour as I have noticed it only serves to make the app seem more inaccurate – particularly when the storm is over 20 minutes away, as there can be several minutes of variation between each update. Using a relative time scale allows the app to fine tune the prediction without the penalty of being seen to be making such changes. Also having tested both I find a relative time scale more useful than the absolute time.
I’ll see if I can add this, but that information is freely available online