My answer is somewhat similar to endolith's but more specific. There was a recent Meta question that was quickly deleted about what was labeled as a 1 character edit. The question in question was Does pbpaste empty the clipboard under some circumstances?Does pbpaste empty the clipboard under some circumstances?, and it makes a perfect example.
The original question included a typo. It read, "The wording seeds odd...". It was probably meant to be, "The wording seems odd...". In the edit, the word was just changed to a different word, "The wording read odd...". That probably would have been better as, "The wording reads odd...", since it refers to an existing document.
Let me suggest a specific rule. The issue with "read" vs. "reads" doesn't affect the usability. The reader may notice that the word is awkward but can keep going without losing the train of thought. That is not the case with "seeds" vs. "seems". These are totally different words and "seeds" makes no sense in the sentence. It causes the reader to re-read the sentence in order to make sense of it and interrupts the process of understanding the question. In this case, the change of a single letter improves the usability of the question.
My suggestion is to make any change if it improves the usability of the answer, even if the total edit is only one character. If it doesn't change the usability, don't bother unless it is part of a collection of changes that together, make the question or answer substantively better.
I would also suggest a context guideline. Some answers don't add much value, and are poorly written on top of that, although they meet the definition of an answer. Other answers provide good information but would benefit from polish. The community benefits when a good, "raw" answer gets improved. Polishing a marginal answer doesn't make it better (the technical term is "putting lipstick on a pig"). Experienced users don't bother editing the marginal answers, but they are low hanging fruit for new users who want to build some rep. We should offer guidance to new users who focus on these and suggest that they focus on where their efforts will provide benefit.