A late discovery?

White to play and win

In the diagram is a simple(?!) study by Van der Heijden from 2002. As with all good studies, the main line contains some non-obvious choices, and solving it might teach you something about King and Pawn endings.
The thing that surprises me most about this study is in fact the date of composition. 2002 seems quite late for a study like this, as I had assumed that most of the tricky king and pawn positions would have been found by the mid 20th century at the latest. But as this study shows, there may still be more discoveries to be had.