Often it isn't harmful, but sometimes it can be, so that is why to be careful of it, I think.
Altering comments (and documentation) is probably not harmful (as long as it remains clear, or in some cases, becomes even more clearly meaning), but altering names of commands and other stuff (especially if it is an exposed API) is more problematic, I think. (Changing them to same words as other things to be confusing, or breaking URLs, or whatever, is also harmful, though.)
I think that it is best to change the vocabulary when the concepts are altered, and to leave it the same when the concepts remain unaltered, although I am not trying to stop you or anyone from writing what they want to write; if they don't like these words, change them, especially if the new words are a better description of it anyways.
(But I believe in freedom of speech: If someone else doesn't want to change their program or document, don't force them to do so. But I also dislike copyright: I believe that if you don't like their program or document or whatever, you should be allowed to make (and use, and distribute) a modified copy.)
* Myer and Sutherlands "On the Design of Display Processors" gives some sense of how we go around in circles (or, hopefully, spirals), but I'm unaware of any reference for how vocabulary changes while the concepts remain unaltered.
An example might be "prefix sum" calculations (or whatever they're called on GPUs nowadays), as these have appeared under different names back to before computers. (when hand calculating, in the days when "computer" was a job description, not a machine, and US Navy calculations took weeks to run, shallow parallel algorithms could save substantial sums)