Foreword: The smart copy/paste/insert functions must be spatially consistent. Speaking "neutrally in 1 dimension": The reference line for pasting + inserting is always the "start" of the selection. Smart Insert: "Before" the selection it inserts void space of the same dimension as the selection. Smart Paste: "Before" the current selection it inserts the cut content. In western cultures in 2-dimensions "before" translates into the following: The upper selection boundary of row-selections (reading from top to bottom), and the left selection boundary of column-selections (reading from the left to the right).
1) Tested build 5633

Smart Cut and Smart Paste now work correctly both for rows and columns.
But sadly for Smart Insert your introduced a new inconsistency which worked fine before:

Having a selection column: Smart Insert inserts the dimensions of the selection at the left selection boundary. Or in other words: Left of the selection it inserts void space of the same dimension as the selection.

Having a selection row: Smart Insert inserts the dimensions of the selection at the bottom selection boundary. This is wrong: It should insert at the top selection boundary. Or in other words: Above the selection it inserts void space of the same dimension as the selection.

If you manage to fix Smart Insert now without affecting Smart Cut and Paste behavior as an unintended side effect, GC finally will have a very intuitive SMART CUT / PASTE /INSERT feature.
2) My proposal of keeping the viewport stable
- as is "content pixel keeps the same Euclidean distance to the canvas before/after manipulations/zooming/image-skipping
- You did not give me any response on this AFAIR. Would this be feasible?
- This would so much improve the comprehension of all growing and shrinking manipulations in the mental spatial model of the user!
- For "Resize Canvas" and "Add/remove Border (Margins)" the viewport stability feature must be designed in a way that the old content pixel is on the same position after the operation, so that you realize which surroundings got added/removed.
- And if users could set the preferred reference-point for zooming and image-skipping (top-left, top-center, top-right, mid-left, mid-center, mid-right, bottom-left, bottom-center, bottom-right)) this would also greatly improve zooming by keyboard with ⌘ +/- and skipping through an image series (Prev/Next image in folder ⌘ ← and ⌘ →)
- And for zooming by gesture I already proposed
Zooming should be towards cursor position whenever the cursor is over the image but got no reaction so far.