If the suggestion is that a single, global visibility or difficulty key allow an annotation of the scale being used, that would all but force data consumers to normalize values into a preferred scale. As @erutan points out, the conversion could be tricky in the absence of more detailed tagging.
To the extent that the conversions are less than straightforward, it could be be useful to tag each feature with up to two keys, one according to a global scale and another according to whatever scale is signposted on the ground or otherwise locally recognized. If a mapper only knows one scale but not the other, they can omit the one they don’t know until someone familiar with the other can make their own observations. This approach would facilitate both automatic heuristics on the part of the data consumer and manual decisions on the part of the end user, albeit with a maintenance overhead as we try to keep the two keys in sync.
A similar bifurcation has been proposed for elevations in the local datum (not just the local units); a variation of it has seen some regional adoption. However, I think the same approach applied to trail visibility or difficulty holds more promise because of the intense interest around these attributes.