What year to use when citing second editions of books

In this case I would say Be Honest: only cite the references that you have read. Even if you have read the "rev 1", you might be (wrongly) citing as "rev 2" a sentence from "rev 1" that was removed in "rev 2"...

It might be obvious, but sloppy mistakes like that happen

If the journal uses the American Psychological Association (APA) style, then you cite the year of the edition and you do not mention earlier editions.

For example, if you are using the 4th edition (published in 1994) of the APA publication manual, you would cite it as (American Psychological Association, 1994) and its bibliographic entry would be:

American Psychological Association. (1994). Publication manual of the American Psychological Association (4th ed.). Washington, DC: Author.

It is always best and safest to cite the sources you've actually used.

If there are two (or more editions) of a book, you generally don't have a complete list of changes from one version to the next. The relevant information might not be in the first edition, or might be outdated from the one you've used. So: use the edition you've actually read.

This is also true if you've used an eprint on the arXiv, but you later find out that it is published (or will be published) in a journal: still use the arXiv reference.

This is subtle: the eprint wasn't ever peer reviewed, and thus is a less authoritative source. Your wording should always reflect that. If then the e-print is published in a journal, and you only change the reference it's like saying you don't care about peer review. It occurs all too often, but it's simply bad science, even if the contents of the two is word-for-word the same. If you change the reference, re-write the relevant sections as well.