Since I've been reading the last of the Holmes canon, I've been looking at questions of the narrative chronology of the Holmes cases -- a very notoriously tangled part of Sherlockiana. A Basic Timeline of Terra 221B by Brad Keefauver is a very good resource, identifying from each case the specific time indicators and giving his own assessment (while comparing it to two more widely recognized assessments). Some of the dates, of course, are quite sure -- Watson explicitly states that "The Illustrious Client" began on September 3, 1902, for instance -- and others have to be pieced together from more fragmentary comments, and yet others are highly speculative, since there is often not much for a chronology to be built on. My favorite of these latter is his attempt to pin down a year for "Silver Blaze" based on the names of the horses, probably closely followed by the even more ingenious (although perhaps a little more convincing) use of Holmes's different attitudes to Turkish baths to fix the relative order of "The Illustrious Client" and "The Disappearance of Lady Frances Carfax".
And, of course, there are the definite problem cases. The most notorious is that of "Wisteria Lodge", in which Watson quite explicitly says that it occurred toward the end of March, 1892. The problem is that Holmes had apparently been killed by Moriarty in "The Final Problem" the year before and Watson doesn't learn he is still alive until "The Empty House" in the spring of 1894. The usual chronologies treat this as a slip of the pen; Keefauver just gives up and suggests that Watson hallucinated the case, although during the time he claimed. Another is "Charles Augustus Milverton", where Watson states that he is concealing the date; yet another is "The Man with the Twisted Lip", in which Watson gets into an argument with someone high on opium about what the date is, and the man with the opium seems to know what he's talking about more than Watson does.
I always find this kind of thing fascinating, providing very interesting examples of how causal reasoning works.