Just got finished coding a study for our meta-analysis. I found two sources from the same authors, one in 1993 (a conference paper) and one from 2000 (a journal article). Turns out they're the exact same paper. So it took them 7 years to publish it. Having gone through the process myself I can believe it in theory, I once had a journal keep a manuscript for over a year--but 7 years from conference proceeding to journal article? Wow. Now, some of the lag has to be due to the authors (perhaps waiting to submit, revisions, etc . . . although the final article looks amazingly identical to the initial conference proceedings). But some blame has got to rest with the existing peer review/publication process as well. I have a hard time believing that the authors sat on this for 7 years.
A couple years back I took a subset of this data (preliminary results of the meta-analysis) and looked at lag between time of data collection and time of publication. The average was 4 years across 57 outcomes (not sure about the number of studies but something < 27). A different benchmark, because this includes time to prepare the manuscript whereas the authors in the study above already had a finished manuscript. And again, part of the time is the authors preparing the manuscript after collecting the data. But all of that said, this still seems to point to journals holding on to manuscripts for a long time. In our field, anything longer than a year from submission to print is probably too long and I'm willing to bet the average is actually longer than that.