+1 on this. We've been working up some images lately where absolute second-to-second sun timing is critical to the view working. We've noticed that the sun just doesn't behave how you'd expect at extreme low-to-horizon altitudes, and this is probably the reason why.
In addition, I think there is an issue with CoronaSun being severely under-powered when very low on the horizon. I can't post any examples due to NDA but it's fairly easy to test this - just load up a city model and corona daylight system in there, and get the sun only just barely poking up above the horizon. You'll find that it's very very low-strength until really quite high. In order to get anything strong enough you end up needing to boost the multi to 2 or more.
Think this model could do with a look. And the age-old request for a 24-hour daylight system still remains... Imagine if we could have a proper, realistic 24-hour corona lighting system that reacts realistically at all times of day including super early/late plus proper transition to blue hour, night, moon light controls........