Don't know how long it takes me, but it's usually from the moment I want up until 3PM or so. I usually get up craving research (when I am on the up and up; other days I hate research and miss the bits and bytes, starting my day in Emacs.)

I rarely read the abstracts. I find interesting papers in the references of other interesting papers.

I usually skip the first 15% of the introductory prose, go to the meat, jump to the conclusions and "future work", and if I think the paper covers enough ground, I fast forward to its references and see if the authors are aware of the seminal works. Only then do I actually read it.

For the truly important ones, you're already aware for their findings from texts, since they're cited often. For the useless ones (the majority) you're just interested in one idea, technique, finding or implementation method, and it's easy to zoom in. But from time to time you will come across a diamond in the rough that isn't as well known as it should.