(From Mikkel:)

It may be helpful to document the preferred approach to animations [at the RenderObject/RenderBox layer].

My current understanding is that when animations (i.e. the lerp values) can be described adequately at the Widget level, that's where animation should be implemented, by arranging for setState on some StatefulWidget to be called at each animation tick.

But sometimes you seem to need to transform between the layouts calculated for your start and end widgets instead. The two cases we have seen involve size-dependent layouts containing animation-relevant information not present at the widget level. That moves animation code into the rendering layer, and it would be nice to have a standard way of accomplishing that.