Not just my opinion, 88% of respondents in this poll right here associate tracktype with several different aspects. That already suggests it functions as a rough classification in practice, rather than a single well-defined dimension.
If tracktype were strictly about firmness only, we likely wouldn’t be seeing so much difficulty agreeing on a precise definition.
As you pointed out, there are many factors involved beyond surface and smoothness, such as firmness, maintenance, engineering, etc. These could also be mapped with more specific tags where needed. That would simply add precision, while tracktype remains a rough classification.
From a data consumer perspective (e.g. routing), this could also be handled through weighting. A router could use tracktype as a coarse signal when it’s the only tag available, but give higher priority to more precise tags such as smoothness, surface, or others when they are present.