1) Do reviewers actually read (or read) articles you cite ?
Sometimes, but not always. When I review a paper, I'll generally check a random subset of citations to make sure the right paper is linked. (It's surprisingly common for references to get out of sync, so that citation  in the manuscript actually should be citing reference  in the bibliography.) I'm often familiar enough with the literature to know if the authors are citing an appropriate paper for common points, so I'll skim over these to make sure everything I already know makes sense. In both cases, I don't actually read the papers, just make sure they pass the sanity check.
If something looks odd (they cite an author supporting something that I didn't think she worked on, or a fact seems reversed, or there's some detail I didn't know about) I will often take a look at the referenced paper. This averages to probably a couple of papers read per paper reviewed, but when I find one oddity I'll often find several, so it's more like some reviews I'll read 4-5 papers, more often I'll just glance at one.
So you shouldn't count on slipping something by reviewers.
As for looking at abstracts only, you can but probably shouldn't. The only time it's relatively safe is when the abstract explicitly states the single point you're using that paper to support ("We found that 72% of Bulgarian tractor repairmen are left-handed"), but even then you may be missing important qualifiers ("We limited our sample to red-haired tractor repairmen with pre-existing arthritis"). As I noted, reviewers may well be familiar enough with the literature to catch these.