Other than extremely minor edits, I think the answer has to be NO.
If I've got the capitalization wrong on a case sensitive variable/function/method/class name in one out of the 20 times I used it in the pasted source code, or if I missed a singular semicolon or parenthesis or closing brace in several, several lines of source code, this is a very minor edit that probably isn't worth the hassle of closing. Any good IDE would have noticed this and depending on the IDE, it may have autocorrected it or would at least be making the correct change recommendation.
But there should never be the slightest amount of guess work in these changes, and they should always be very minor.
We disallow and rollback edits to indentation, because that is a reviewable aspect of the code.
If the change is significant as changing a function call (in the example, we changed between
.append and assignment, well now you're significantly changing the actual source code.
If the original code works, you need to post this change as an answer explain why one is better than the other--just like we require you to comment on why good indentation is good rather than allowing users to just fix it in the question. If the original code does not work, then we're probably applying a slight bit of guess work to make the code work, and we may have made a change in an edit that the original asker doesn't understand.
In answers we post, we generally feel the need to explain the changes we recommend so that the user understands why our recommendation is better than what he has. If a question's source code has been edited only by the original asker, it is fair enough (though not necessarily always true) to expect the original asker to understand why his code does what it does, and so our answers' explanations can be tailored to a user with that level of understand. If someone comes in behind the asker and changes the source code in the question, we can no longer make the assumption that the asker understands the source code in the question.
If the code does not work, the asker needs to fix it themselves or post the exact same source code as a question on StackOverflow.
For clarity, trivially minor edits that require no guess work, and arguably, also do not require reading the plain-English part of the question should be okay. These would be things like an obvious copy & paste that missed a few characters.
Anything more significant than this, anything that requires guess work or reading the plain-English part, anything that changes which function was called or which class was used, these are all things that should be rolled back and only be changed by the asker.
In the specific case offered as example in the question, I think the rollback was appropriate.
The rule should be no, these edits are not allowed. But exceptions should be made for trivially minor edits which require no guess-work.