In the U.S., where the term originated, “sidewalk” is usually also defined in highway codes, as is “driveway”. However, I think the vast majority of mappers here prefer to apply the tag intuitively based on the geometry and appearance, because the legal definitions are generally intended to model the intuitive understanding. To the extent that the definitions differ, mappers here would consider any rule lawyering to be pedantry.
cycleway=* answers the question, “What kind of cycling infrastructure is it?” Thus, cycleway=track means it’s a cycle track, not that cyclists must ride on a highway=track. To me, sidewalk is inaccurate, saying that there is a sidewalk that is a kind of cycling infrastructure. Others apparently read it as, “This is a sidewalk for bikes,” or, “This is a bike path where there would ordinarily be a sidewalk.”
From a motorist perspective, it’s all the same anyways. But to me as a mapper, this conflation is problematic because there can also be cycling “infrastructure” in the form of a sidewalk where a sidewalk would normally be – the no-build option so beloved by American highway planners. These are sidewalks but not anything recognizable as bikeways out of context.
Practically speaking, a renderer might have a heuristic to hide sidewalks because they’re redundant to streets, or a router could tell a pedestrian to follow [name of street] instead of “unnamed sidewalk”. Replacing cycleway=sidewalk with something different like cycleway=sidepath or, ugh, is_sidepath=yes would exclude these sidepaths from such heuristics, but I think that’s a good thing. These are more substantial paths replete with their own street furniture that doesn’t belong to the parallel street.