Yes! Spot On! You don't need to leave the functional modeling space and move to the Signal Integrity space before the misuse of the global command bites you!
I was sitting here thinking about how to "hack my design" to overcome a lack of forethought, strongly considering my occasional misuse of the global command when I read your post. You talked me out of it!
At least with the .global command, the end user can edit the files (assuming nonencrypted files) and after much tribulation can get files that work together.
If you have one buffer model that used scale=1u and another buffer model that didn't use scaling, then you're pretty well stuck. I got in that situation some years ago and attempted to find complete documentation on what model parameters are affected by the scale option. I couldn't even find where it was documented to that detail, so trying to edit the models, even with a script, was impossible.
Whenever I talk to people about SPICE models, I try to always throw in the counsel to NEVER use scale option, and RUN AWAY from .global statements. Please!