Must Research Talks present Work in Progress?

I gather the norm is that research talks (colloquia, etc.) are meant for presenting unpublished work; work that is, at least nominally, "in progress".  But is there a good reason for this norm?  Just off the top of my head, I would have thought that research talks served two main purposes: (i) feedback, and (ii) dissemination.  Moreover, "read in advance" workshop-style events aside, I expect the main benefits for all involved stem from the latter: the audience gets exposed to (hopefully interesting) new ideas, and the speaker gets to disseminate her ideas, perhaps build up her academic reputation slightly by becoming better known to the audience members, etc.

And while opportunities for feedback are no longer such a priority for published work (though it surely never hurts to hear new objections, etc.), I would think the benefits of dissemination would be all the greater when it comes to presenting one's published work, as selection effects mean it is likely to be of higher quality than one's current work-in-progress. The audience would benefit more from being exposed to your most interesting ideas (assuming you aren't so famous that they'd heard it all before), and you too would presumably benefit more from disseminating your best ideas rather than simply your most recent ones.

So, why doesn't this happen? (Or, if it sometimes happens, just without my being aware of it: why not more often?)