How to handle plagiarism on method that does not affect outcome results when reviewing a paper?
There is no such thing as unimportant plagiarism. And three to four stolen paragraphs is not small.
You did the right thing to report it to the editor. But I also would have rejected the paper. There is no place in academia for academic misconduct. It certainly shouldn't be published.
It is not for a reviewer to decide whether the plagiarism has actually happened and whether the paper has to be rejected because of it. This is the editor's call. However, it is reviewer's duty to note the similarity and express concerns about possible plagiarism to the editor, supported by evidence.
Plagiarism is a serious breach of academic integrity and should not be tolerated. The amount of the copied text is not really relevant here: authors should not pass someone else's words and work as their own, however big or small it is. If the text was borrowed, it should've been properly attributed. This is not about the validity of result, but about the principles on which the academic community stands.
Regarding unimportant plagiarism, I have to note that sometimes plagiarism is unintentional. Last year, we submited a paper to a top-tier conference, where we accidentally plagiarised two sentences from a paper of the program chair (who we were pretty sure that were also one of the reviewers). The paper was still accepted to be published.
That program chair published a well-known paper 10 years ago, which formalized a model, and proposed a naive algorithm (brute-force) to compute some entities on this model. The approach was demonstrated on some toy programs, written in a toy language, with a couple lines of code.
We were the first, in 10 years, to propose a practical algorithm for this model, which scales to thousands of line of Java. Of course, all of us read that paper countless times, and discussed it for several months. As a result, many of its sentences stuck in our heads, and somehow made their way to our paper. Notably, two sentences were exactly identical, since they described the settings for the problem.
One of the reviewers, whose review only appeared after the rebuttal phase, explicitly said that he did not read technical details, and only wanted to give editorial comments. So we were almost sure that he was the program chair.
That reviewer was extremely upset about our discussion in the related work, since we only compared our algorithm with the naive brute-force, and he felt we did not give credit to the model that we implemented (we formalized the model in a different way). In particular, he pointed out those two sentences that we lifted verbatim from that well-known paper, as a proof of the influence of his paper to ours.
But that didn't result in a rejection, and it was not a conditional acceptance, i.e. they didn't review our paper again.