I noticed this spell is a bit of a pain to DM, so I am curious how other people have done it.
When I first ran it I considered all the figments to be in the same space as the original creature; which was simple to adjudicate, but not exactly to the letter of the rules.
Before I ran the spell in the next session, a player pointed out that images could be up to 5' from each other. Hrm. OK, so I made up some "image" counters and sprinkled them on the board so each image had a tangible position.
But this caused people to be able to deduce which image was real by logic ("he hit me and then moved, so assuming he moves 30' then he's probably that one but certainly not that other one..."). It also caused the images to condense down whenever the creature did something observable, like attack (ie, all images of an attacking creature must be within 5' of the victim).
So, how do people physically DM this spell? I'm tempted to go back to my original method, since it was so much simpler. Both methods played out roughly the same, even in the case of area of effect spells (somewhat to my surprise.)
----
BTW, I played that once the "real" image is discovered he may be attacked freely by whoever discovered him for that round; but any additional attackers must still find out for themselves.