That is all true, but I do not consider it a valid argument.
At the moment KiCad looks at the date stamp at the beginning of a file, and if it’s in the future it just refuses to open the file. This is “safe” and simple,but it is also very dumb.
Having an import filter that just dumps everything it does not understand in a text window (with a save option) puts the user in control. If the older version does not support rectangular pads with rounded corners, then it’s easy enough to replace some footprints with whatever that older version does support. If a few circles are missing, so what? As long as you have some Idea of what is missing, then it’s very likely easy to repair manually. Being able to port 95% is much better then KiCad just giving a big finger to the user, especially if the other 5% is dumped in a text file, so you know what is missing and where you need to go to fix issues.
Some time ago I wrote:
And it’s a valid and useful method. The concept of a “footprint” does not exist in a gerber file at all, but it’s easy enough to just delete all the pads, and then place a new footprint on top of the existing track ends. With a back import from Gerbers you get about 80% to 90% of the PCB from any project which has gerbers available. Reconstructing the schematic was a lot more work. It was mostly a manual re-creation from a .PDF file, but by updating the PCB regularly and DRC which flags connection errors it’s still much better then “nothing”. I made a few mistakes by re-creating a KiCad schematic from the pdf, and those were caught by the DRC because all tracks were preserved from the Gerber file, and they implicitly define the netlist.