I assume MW uses a process similar to the
Oxford English Dictionary. The key that tends to produce weird lags is that dictionaries usually only care about usage in print so slang tends to lag well behind since it is much less likely to get into print.
Also, consistent usage over time is a factor. MW lists their first discovered use of "unibrow" as 1988 but it may be that it splashed onto the scene and then disappeared from print sources for most of the next decade.
The example I find interesting is "ringtone" which is dated to 1983 but I imagine until the last couple of generations of cellphones was purely an industry technical term.