Feature detection is not the only reason the new protocol exists; it offers a variety of options to create scaled text (not only x2) and alignment, and it can be detected reliably, unlike the older sequence which is also not widely supported anyway.
This was my experience when writing the tool, which had started out only with the image rendering, but then could use the new kitty protocol. The old sequence is moot for this use case, because it's impossible to query if supported and only can scale to x2.
Also the tool you shared has nothing to do with this. It’s just a 3rd party utility.