If he had done this, the blocksize limit would be set at 11.4 MB now and I bet everyone would be pretty happy. It would just be taken for granted that "of course the limit grows with time because the systems grows with time too!"
More likely the 1 MB would have started at a much lower number as 1 MB was both overkill for near-term usage and hard to support with current technology at the time. 1 MB was clearly chosen with an expectation that it wouldn't be changed very soon. So given an autoscaling limit, it might well have been something more like 50K + 50%/year, giving us approximately 600K now.
FWIW, I think 600K would still be a reasonable limit for the present (if that were in effect we'd likely see better fee management tooling), but I also think something like 2-3 MB is reasonable, or maybe 1 MB plus 50% per year (or a tapering growth rate or something). As you say getting it exactly right is difficult and unnecessary.