Just because you are an editor does not mean that you have to deny your friendships. This is the time where you can payback the favours to your colleagues. And always remember that you have the final decisive power of acceptance. It is perfectly acceptable to overrule the reviewers of a paper to get a paper of another big-shot friend accepted; most likely, the paper was reviewed by a student anyway.
Once you are formally an editor and your name is on the cover of the journal, resign from the journal on a matter of high principle. Then go to the press. This will make you more famous.
Do not feel obliged to process papers promptly. There are several reasons.
Authors will realise you are a big-shot and thus very busy.
You should always sit on a paper for a year or two to let it mature before passing it on; no-one likes to review under-ripe papers. This is particularly useful when you are working in a similar area as you can ensure your work is published first. Furthermore you will be able to force the authors of the submission to cite your work as prior art.
A hallmark of a top journal is that it, like an exclusive nightclub, has a long queue of wannabes waiting to be admitted. Remember your role as the gatekeeper is to keep them out!
Keep your competitors unproductive by subjecting them to an endless stream of pedantic revisions that are time-consuming and frustrating to implement. For example, it is perfectly acceptable to request the authors to answer some of the open problems that they state. Another good trick is to require them to move every comma in the paper (this can be done repeatedly over several re-submissions). Temporal consistency is not necessary.
Keep the author in suspense by not answering any queries; if he is not prepared to wait, then he is not a worthy member of the club.
Maintain a secret and complex set of rules for acceptance; you are editing a machine learning journal and if the authors of submitted papers can not inductively learn the rules they are not worthy of publishing in your journal.
If there is a paper submitted you do not like, but looks solid to most reasonable people, you can nevertheless get it rejected by picking inconsistent sets of referees. For example, send it to a Bayesian and a frequentist, and reject on the basis that it must be badly written as the expert referees can not even agree amongst themselves what the contribution is.
You were appointed an editor because of your objectivity; conflict of interest is something that applies to other people. For example, you can reasonably accept all papers from your institute because it is of such high quality by definition that you can be sure the paper is too.
Make sure you always weave into any conversation that you have to go to the editorial board meeting. For example, whenever someone tells you of some exciting new advance, you can say, in a casual tone, “Yes, that caused quite a stir at the NIPS program committee.”