What would the impact of using a different edge detector/combining features be?

Probably different edges. For some fun, try looking at the results of using canny edge detection. Always thought it looked like strings/noodles.

How might we go about finding diagonal edges? It looks like we might need a bigger kernel, and also the matrix is square, so would this alter the functionality of a diagonal edge detector? Would a diagonal edge detector be any worse at detecting edges because of a square matrix that is more intuitive to use for vertical and horizontal edges?