Are studies created or discovered?

A while ago I saw part of a discussion on using computers to compose problems and studies. While I didn't read the whole discussion, and indeed I've now forgotten where I read it, there was one point I found interesting. It was the suggestion that studies aren't composed, as all possible chess positions already exist. Instead all that happens is that the correct evaluation of an existing position is found.

While this seems to have an undeniable logic about it, it of course doesn't explain why some problems are more highly regarded than others. This is because while all positions may exists, some positions are more interesting than others. As an example compare the following study by Grigoriev against the well known position of wKe6,wPe5,bKe8. While both have the same evaluation (White wins), the Grigoriev position poses a greater challenge.