TL;DR
By all means, invest time in improving questions and answers on the site whenever you come across a post that needs improvement. However, actively seeking out edits that are primarily "missions of grammar" or otherwise trivial edits can waste other people's time, and may color their perception of the editor's activities on the site. You don't want people thinking you're rep-farming or badge-seeking in your edits, or simply being a philosophunculist or grammar nerd and wasting their time in the process.
Don't be afraid to make necessary changes. On the other hand, if one starts getting a lot of rejected edits or rollbacks, it's probably a good idea to reconsider whether the community values the changes being made.
Edits Should Make Posts Better
In general, the goal of edits should be to:
- Make a post or question title clearer.
- Fix outdated links or other "broken window" effects.
- Help bring a problematic post into on-topic status.
- Clean up a post that violates our community standards (e.g. removing signature lines).
Aside from creating noise, trivial edits are those that don't add significant value. Defining "value" is hard, because it varies somewhat across SE communities. In general, though, I would classify trivial edits as:
- Edits that fix minor spelling errors or insignificant typos while leaving poor grammar or other issues behind.
- Edits that focus on semantically minor word-choice selections.
- Edits that attempt to change the authorial voice without material improvement to the content of the post.
- Purely stylistic changes that add no real value, e.g. deciding that someone's post should use
fixed-width instead of italics when the author's usage wasn't unclear.
- Highly-upvoted questions are probably fine just the way they are.
- Accepted, highly-upvoted answers should most likely be left alone.
In addition, edits that change authorial intent (other than to bring an otherwise off-topic or non-conforming post into line with community standards) are never okay.
Perception Matters
Back in the day, too many trivial edits would also trigger an automatic conversion of a post to community wiki status. This was often undesireable. That doesn't happen anymore, but that legacy often informs the opinions of people reviewing your reviews.
Another consideration is that trivial edits are often perceived as either unnecessarily critical (which violates the "be friendly" code of conduct), or uselessly punctilious. If your edit doesn't add enough value to overcome either of these perceptions, I'd recommend that you skip it.
As an example, while I would personally like to edit EVERY...SINGLE...POST that fails to include the Oxford comma in lists, or replace commas with semicolons in lists with conjunctions, doing that throughout the site as a personal mission wouldn't accomplish much except to annoy people and clutter up the home page. However, I will certainly fix those types of issues if I'm making other, more significant changes to a post anyway.
Consider that until you reach a certain level of reputation, your edits enter a review queue. Cluttering up the queue with edits that other reviewers don't consider meaningful or useful can result in rejected edits and a more jaundiced view of your future edits. While people should review proposed edits, flags, and other actions idempotently and without regard to the person making the change, realistically someone faced with dozens of "useless" edits from the same person is likely to start rejecting that individual's queued or future edits without giving them the benefit of the doubt each time.
People's time also matters. Reviewers will general spend only a couple of seconds on each item in the review queue, and so an edit that only adds one comma or capitalizes a single word is likely to be rejected as trivial most of the time. Even if it's really an "improvement" like changing ruby to Ruby deep inside a long post, it's unlikely to be enough of an improvement to warrant wasting 4+ people's time. Most edits by users with less than 1,000 reputation (2,000 on non-beta sites) involve the editor, two or more non-diamond reviewers, a message to the original poster, and sometimes the moderator queues. That's a lot of people to involve in something that others will consider trivial, and seconds can multiply into man-hours once you start considering the issue at scale.
If we correct all problems that we observe and they are very few (let's say 1 or two words), is that considered trivia and, therefore, discouraged?Usually such edits should be discourage, but there can be exceptions where such minor edits make a material difference. There's a network effect to edits that I address in more detail in my answer below.