Sounds like a re-occurring issue with understanding terminology used by many people.
"Non-destructive" is often understood as having the ability to do steps A, B, C, then return to whatever changes/settings were applied at step A and being able to modify them withing having to redo steps B and C, and still get the same image as if you would have re-done steps B and C.
You might say that "non-destructive" is not a good name for this ability, but you should understand that it is viewed as a benefit by many. Since Affinity obviously aspires to be non-destructive in this sense with the many live adjustments, so it makes sense to point out that neither of the solutions presented can achieve that.
Consider step A to be the creation of the shape, step B being the creation of blur layer, and step C being the copy of the shape to be also a mask of the blur layer. You can not change the parameters or type of shape (step A) and get a consistent result, without also repeating step C and copy the new shape to be mask of the blur.
Edit: Note that I am assuming that one would like to reproduce the original Background Blur, not just a blur masked to a shape. In proper background blur you have a shape/image with it's own colors and when that is blended in, the background color used for blending is blurred before blending. See the animation with the watch earlier where the layer with background blur has a darker gray color and text as well.