It always help to have concrete examples of the issue being discussed in a meta question, if that doesn't put any of the users involved in a difficult situation. Since I don't think this is the case, I'll go out on a limb and hazard a guess. Are you talking about these edits that came out today?
If so, I happened to come across a few of these myself as well. I approved most of them, but skipped this one that I was unsure of. I would have rejected it, but I now think 4 times before putting down decent contributions. In the end, two other reviewers rejected the edits, so we are now very much certain that the rejection was indeed needed.
In this particular case, the user is really just trying to be helpful and the change in meaning are actual improvements, and are harmless. In any case, the edits can always be ultimately checked by the OP anyway, so I don't this particular case is cause for concern at all.
In general, as an editor, feel free to reject the edits you think change the original meaning too much, making sure you select the correct reject reason. If needed, choose the custom reason and explain very well why the edit is being rejected. That has a very good didactic effect on the editor and will probably fix any behavioral issues. If it doesn't, flag the question for moderation action or bring the issue to meta, just like you did.
Just be careful not to scare the editor away. These contributions are valued here and we want these sort of people around. Over time I learned that the best way to correct minor problems is being gentle. Being too hands-on usually provokes unwanted commotion.