Exactly the comment I was looking for - I've done a bit of work classifying time series using 1-NN and DTW as the distance metric and it works quite well.
I was under the impression this work was more well-known, though. It's surprising to not see it mentioned considering how long DTW has been around.
(he does have interesting stuff here: http://www.cs.ucr.edu/~eamonn/selected_publications.htm)