Motivation and justification for editing
It's important to note that, as described on the tour for each site, Stack Exchange sites are not discussion forums. Unlike in a forum environment, which enables a conversation between someone asking a question and an expert who can answer it, the goal of this model is to produce useful reference material that can outlast the original question - for as long as asking it remains valid. The original incarnation of Stack Overflow was created specifically due to frustration with two main experiences:
Trying to research a programming topic, finding posts on a forum, and then trying to follow a long chain of replies that might ultimately lead nowhere in order to answer one's own question.
Trying to help people on a forum, only to discover that people overwhelmingly either can't clearly describe the problem, repeatedly ask about the same basic topics in slightly different ways, or both.
Of course, most of these problems are addressed by enforcing standards for questions (by closing those which don't meet standards and voting to rate them for quality) and by closing duplicates (external search engines don't care about post score, but they do care about inbound links; and in some cases people are automatically redirected).
But aside from determining the best way to phrase (and scope, and delineate, and illustrate...) a question, and the best answers for that question, we want to improve the reader's experience as much as possible. Posts on Stack Exchange are licensed publicly under Creative Commons because they are supposed to exist for everyone's benefit. In general, the people who read a Q&A greatly outnumber the people who participate in creating and editing it. They're coming to Stack Exchange because they, too, have a question; and respecting and conserving their time is a priority. Readers should be able to:
Determine as quickly as possible that they have found the right question;
Understand a clear problem statement or specification, as a backdrop for understanding the answers;
Read answers that get right to the point for simple problems, and are appropriately organized for more complex ones; and
Feel like they're reading polished reference material, rather than eavesdropping on a conversation.
Therefore, aside from simply fixing clear grammar/punctuation/formatting issues, editors need to concern themselves with:
"Noise"
Probably the most common reason for edits that get complaints from the original author is that the post was judged "noisy". The most easily understood forms of noise are greetings and signatures. But more generally, we want to exclude all conversational language from questions and answers, along with anything that is really about you as opposed to the question or answer. In particular, we only care about your motivation for asking a question, insofar as it helps to understand exactly what the question is.
This is also covered by the "Expected behavior" Help Center article, but not in very much detail. You may also want to read What should I keep out of my posts and titles?.
Removing unnecessary detail
A good question will, almost always, naturally require more words to answer properly than to ask properly. The reader's attention span is limited, and questions are expected to have a reasonable scope so that they can be answered coherently.
Thus: as a rule of thumb, a well asked question should fit comfortably on your screen without scrolling, unless explaining the problem really requires showing a large image or including long technical details about how to reproduce an issue.
In particular, it is generally not helpful for a question about how to perform some simple task, to include details about what the OP has already tried. If the goal is to figure out what went wrong, that's different from a how-to question. Please don't try to ask both at once. Besides which, "what went wrong" might be completely unrelated to what you're trying to do.
Many of the best, most valuable questions on a Stack Exchange site - for example, programming questions on Stack Overflow about simple tasks that every user of a programming language needs to understand, which get thousands of upvotes and millions of views and hundreds of duplicate links - are a single sentence long.
Similarly for answers: content which goes beyond the scope of what was actually asked should generally be removed, even if it's accurate and informative. In many cases, seemingly useful background information is really the answer to a separate question. If it hasn't already been asked, you can ask and answer it yourself, and link to that Q&A from another answer. Refactoring information like this makes it easier for the reader to choose whether to skip over the details.
Advanced formatting and copy-editing considerations
Since they should be short (as explained above), as a rule of thumb questions should generally not have section headings. Consider that if you added section headings to a question that didn't already have them, you would be adding vertical space for the headings themselves, and probably also increasing the total length of the prose. Having e.g. a "background" or "motivation" section on a question also just invites noise that doesn't belong in the first place.
In many cases, questions and answers can be made shorter and easier to read by integrating information into the flow of the text, rather than having bullet-point lists everywhere. When you write a post, before reaching for the * or - keys, consider exactly what you hope to accomplish with that formatting. For example, a list of external reference links can often be replaced with comma-separated links in the middle of sentence with no loss of clarity.
Finally, just because text formatting is used correctly doesn't mean it's used optimally. Consider whether judicious use of boldface and italic could make the post easier to read quickly. In particular, consider using boldface to highlight an "outline" of a post - such that reading only the emboldened words gives the reader a general idea of the post while skipping over detailed explanations.
How to judge edits
There's no clear, objective rule metric for edits that makes much sense. Opinions often vary about whether any given edit is "trivial" (try searching on per-site meta for "trivial edit" to get a sense of it), and whether that should be disqualifying in a given situation. But as some rules of thumb:
Single-word changes etc. will generally be accepted if they're objectively fixing a spelling mistake (or, say, the conjugation of a verb) or removing needless profanity. However, editors should at least scan through the rest of the post and correct any other such errors they notice all at once.
Users with unilateral editing privileges (generally, 2000 or more reputation) are typically given more latitude in editing, especially on Stack Overflow. Edits from lower-rep users have to be reviewed (and there's a small rep bonus on the line), which drains the community's resources; and such edits are put in a queue which can block other edits site-wide when full. Since users with the editing privilege can bypass these restrictions, and are expected to understand editing policy, there's little reason for them to hold back from small changes where necessary.
If it isn't clear whether something is an improvement, imagine making the opposite change. Does that seem like making the post worse?
Of course, clear vandalism should be rolled back immediately. But if you don't like how your post was edited simply because it tramples on your personal style, please first take a deep breath and try to see things from the editor's point of view. Assume good faith, check the edit summary if applicable, and be kind and charitable. If you still aren't convinced that the edit improves your post, consider taking the matter to per-site Meta rather than just rolling the edit back.
