John Miller is joining the beta test team, and so for 10.4 we'll be able to iterate this to resolution. I'm nearly certain that it has to do with de-optimizations we did in 10.3 for which caused faster queries of new map data. I believe that if I just undid these few changes, it would probably return to good performance for John. So then, the trick becomes on how to make iFly adaptive per device, so that it doesn't cause choppiness on ANY devices. This fix should make it in for 10.4, and hopefully John Miller will help us validate it.