I finally sat down to do some tests with 0402 parts using standard 8mm (4mm advance) reels. You can set the Indx value to 0.5 which will cause the reel to advance every second part. I haven't tested it extensively, but I can confirm that it seems to work reasonably well and the placement looked accurate on the half-dozen 0.1uF 0402 caps I placed (using an 0603 pad). Apparently you might get some misses if the second part pops out of the tape, but I'm guessing the miss rate is still pretty acceptable and the fact that 0402 parts are much cheaper makes up for it. It's 11€ for a 10K reel of 0.1uF caps versus 15€ for a similar 5K reel of 0603s, so even if you have a very high 20% miss rate (I didn't get any misses myself) you're still way ahead saving both PCB space and money without having to use dedicated reels. Apparently 0201 is possible with the same approach, though I need to order a special nozzle first to try it out. I'm guessing the miss rate will be higher, but when I get the nozzle I'll post some feedback here.
I tried placing some BGA parts as well (LPC3131, which is a 0.8mm pitch 180 ball part 5 rows deep) and the placement looked pretty accurate. It took a few tries to get the vision setup properly, but once you get the part definition right it looks pretty dependable. Curiously, the hard parts give me less trouble than the discretes with the MDC machine. I need to try with some real PCBs, though, and reflowing to see if I get any bridges, etc. Again ... I'll post some results in a month or so when I have more of a chance to try with real boards and the right footprints, but I'm fairly optimistic after my first attempts.
I've attached an image of the 0402 caps with the Indx set to 0.5, where you can see two part highlights rather than the usual one.